娀 Academy of Management Journal

2011, Vol. 54, No. 2, 393–420.

BUILDING BETTER CAUSAL THEORIES: A FUZZY SET

APPROACH TO TYPOLOGIES IN ORGANIZATION RESEARCH
PEER C. FISS
University of Southern California

Typologies are an important way of organizing the complex cause-effect relationships

that are key building blocks of the strategy and organization literatures. Here, I
develop a novel theoretical perspective on causal core and periphery, which is based
on how elements of a configuration are connected to outcomes. Using data on high-
technology firms, I empirically investigate configurations based on the Miles and Snow
typology using fuzzy set qualitative comparative analysis (fsQCA). My findings show
how the theoretical perspective developed here allows for a detailed analysis of causal
core, periphery, and asymmetry, shifting the focus to midrange theories of causal
processes.

Types and typologies are ubiquitous, both in every-
day social life and in the language of the social
sciences. Everybody uses them, but almost no one
pays any attention to the nature of their
construction.
-McKinney (1969: 4)
The notion of causality plays a key role in both
the strategy and organization literatures. For in-
stance, cause-effect relationships are the central
way in which strategic decisions and organization-
al structures are understood and communicated in
organizations (Ford, 1985; Huff, 1990; Huff & Jen-
kins, 2001). Building on this insight, the cognitive
strategy literature has aimed to map and explain
the causal reasoning of managers regarding both
organizational performance and competitive envi-
ronments (e.g., Barr, Stimpert, & Huff, 1992; Nad-
karni & Narayanan, 2007a, 2007b; Reger & Huff,
I would like to thank Associate Editor Dave Ketchen
and the three anonymous reviewers, as well as Paul
Adler, Harold Doty, Omar El Sawy, Ferdinand Jaspers,
Mark Kennedy, Alan Meyer, Phil More, Willie Ocasio,
Charles Ragin, Nandini Rajagopalan, Chuck Snow, and
seminar participants at Cambridge University, UC Irvine,
IESE, HEC Paris, the University of Illinois at Urbana-
Champaign, the Rotterdam School of Management, the
University of Southern California, and the Wharton
School, for their helpful comments on the ideas ex-
pressed here. Parts of this research were presented at the
24th EGOS Colloquium in Amsterdam and the 2009 Or-
ganization Science and Utah-BYU Strategy winter con-
ferences. I gratefully acknowledge financial support from
the Social Sciences and Humanities Research Council of
Canada and from the Office of the Provost and the Grant
Program for Advancing Scholarship in the Humanities
and Social Sciences at the University of Southern
California.
393
Copyright of the Academy of Management, all rights reserved. Contents may not be copied, emailed, posted to a listserv, or otherwise transmitted without the copyright holder’s express
written permission. Users may print, download or email articles for individual use only.
1993). Similarly, cause-effect relationships are the
main building blocks of the organizational design
literature and have recently received increasing at-
tention (e.g., Burton & Obel, 2004; Grandori & Fur-
nari, 2008; Romme, 2003; Van Aken, 2005).
A key way of organizing complex webs of cause-
effect relationships into coherent accounts is by
means of typologies. As Doty and Glick (1994) ar-
gued, typologies are a unique form of theory build-
ing in that they are complex theories that describe
the causal relationships of contextual, structural,
and strategic factors, thus offering configurations
that can be used to predict variance in an outcome
of interest. As such, typologies have been very pop-
ular and form a central pillar of both the strategic
management and organizational literatures. For in-
stance, typologies such as those of Blau and Scott
(1962), Burns and Stalker (1961), Etzioni (1961),
Miles and Snow (1978), Mintzberg (1983), Porter
(1980), and others have figured prominently in
both fields of research and continue to draw con-
siderable attention (e.g., DeSarbo, Di Benedetto,
Song, & Sinha, 2005; Kabanoff & Brown, 2008;
Meyer, Tsui, & Hinings, 1993).
Typologies are theoretically attractive for a num-
ber of reasons. Because of their multidimensional
nature, the configurational arguments embedded in
typologies acknowledge the complex and interde-
pendent nature of organizations, in which fit and
competitive advantage frequently rest not on a sin-
gle attribute but instead on the relationships and
complementarities between multiple characteris-
tics (e.g., Burton & Obel, 2004; Miller, 1996; Sig-
gelkow, 2002). As such, typologies at their best
result in integrative theories that account for mul-
tiple causal relationships linking structure, strat-
egy, and environment (Child, 1972; McPhee & Scott
394 Academy of Management Journal
Poole, 2001). Typologies are furthermore helpful
because they provide “a form of social scientific
shorthand” (Ragin, 1987: 149) for these multiple
causal relationships by simplifying them into a few
typified and easy-to-remember profiles or Gestalten
(McPhee & Scott Poole, 2001), inviting their use as
heuristic tools for researchers and practitioners
alike (Mintzberg, 1979).
However, for all of their theoretical attractiveness
and considerable success in portraying cause-effect
relationships, typologies also present considerable
challenges. The same features that make typologies
so appealing to researchers and practitioners—
namely, their holistic approach and their ability to
combine complexity with parsimony—are at the
same time some of the greatest disadvantages of
configurational arguments. Perhaps most impor-
tantly in this regard, typologies tend to be based on
a logic of consistency; that is, they are usually based
on the notion of “fit” among the different parts that
make up the overall ideal type or configuration.
Although a few researchers have pointed to the
varying theoretical importance of constructs in ty-
pologies (e.g., Doty & Glick, 1994; Mintzberg, 1979),
the proponents of most prior typologies have ne-
glected such considerations by asking scholars to
accept them in toto. However, this is a problematic
proposition; although a holistic approach can be
useful at times, theorizing is more likely to end
once a typology is identified, thus limiting under-
standing as to what causal mechanisms are at work
and what is driving an effect (McPhee & Scott
Poole, 2001; Reynolds, 1971).
In this study, I argue that the logic of consistency
that flows from the holistic nature of typologies
presents a fundamentally problematic assumption
that is likely to lead both researchers and practitio-
ners astray. For example, existing typologies are
likely to contain inconsistencies, trade-offs, and—
perhaps most importantly—irrelevant elements.
However, if not all parts of a configuration are
equally important, the issue becomes this: Which
are the critical aspects in a typology, and which
elements are nonessential? The challenge of typol-
ogies thus is determining what really matters (and
to what degree) in understanding the causal struc-
ture of a type.
How, then, can one reconceptualize the causal
relationships in typologies to move away from a
fully holistic view to a finer-grained understanding
of what is causally relevant? To address this prob-
lem, I develop a different theoretical perspective on
causal relationships in typologies. Drawing on ar-
guments from the strategy and organizational de-
sign literatures (e.g., Grandori & Furnari, 2008; Sig-
gelkow, 2002), I argue that typologies frequently
April
consist of a core and a periphery, with the core
elements being essential and the peripheral ele-
ments being less important and perhaps even ex-
pendable or exchangeable (e.g., Hannan, Burton, &
Baron, 1996). Specifically, I develop a definition of
“coreness” based on causal connection to the out-
come of interest. Accordingly, I define core ele-
ments as those causal conditions for which the
evidence indicates a strong causal relationship
with the outcome of interest and peripheral ele-
ments as those for which the evidence for a causal
relationship with the outcome is weaker.
Adopting the theoretical perspective that I pro-
pose here contributes to the prior literature in sev-
eral ways. First, the distinction between causal core
and periphery allows me to extend prior theory by
introducing the notion of neutral permutations.
This notion suggests that, within a given typology,
more than one constellation of different peripheral
causes may surround the core causal condition,
with these permutations of peripheral elements be-
ing equally effective regarding performance. As I
will show, this notion extends current theories of
equifinality, the idea that “a system can reach the
same final state from different initial conditions
and by a variety of different paths” (Katz & Kahn,
1978: 30). Equifinality has recently received in-
creasing attention in the management literature
(e.g., Doty, Glick, & Huber 1993; Fiss, 2007; Gresov
& Drazin, 1997; Marlin, Ketchen, & Lamont, 2007;
Payne, 2006) because it provides a theoretical un-
derpinning for the persistence of a variety of design
choices that can all lead to a desired outcome, thus
offering considerable promise for organization the-
ory (e.g., Ashmos & Huber, 1987; Short, Payne, &
Ketchen, 2008).
Second, the notion of causal core and periphery
extends prior thinking on cause-effect relationships
by implying causal asymmetry (Ragin, 2008)—that
is, the idea that the causes leading to the presence
of an outcome of interest may be quite different
from those leading to the absence of the outcome.
This view stands in contrast to the common corre-
lational understanding of causality, in which
causal symmetry is assumed because correlations
are by their very nature symmetric; for example, if
one models the inverse of high performance, then
the results of a correlational analysis are un-
changed except for the sign of the coefficients.
However, a causal understanding of necessary and
sufficient conditions is causally asymmetric. That
is, the set of causal conditions leading to the pres-
ence of the outcome may frequently be different
from the set of conditions leading to the absence of
the outcome. Shifting to a causal, core-periphery
view of typologies allows for such differing sets of
2011
causal conditions to exist across the range of an
outcome, with one set leading, for instance, to av-
erage performance; a different set, to high perfor-
mance; and yet another set, to very high perfor-
mance. As I argue, making this theoretical shift has
important implications for understanding the rela-
tionships among strategy, organizational design,
and environmental context, and thus it carries sig-
nificant implications for organizational design and
strategy more broadly.
I empirically tested the arguments developed
here on a sample of high-technology firms using a
novel methodology for modeling causal relations:
fuzzy set qualitative comparative analysis (Ragin,
2000, 2008). This approach is based on the idea that
causal relations are frequently better understood in
terms of set-theoretic relations rather than correla-
tions (Fiss, 2007; Ragin, 1987, 2000, 2008; Ragin &
Fiss, 2008). Drawing on the well-known Miles and
Snow (1978, 2003) typology as an example, my
results indicate that the core-periphery model of
typologies proposed here offers both a different and
a finer-grained understanding of the causal rela-
tionships among the elements of typologies, thus
addressing a key issue for both strategy and organ-
izational design (e.g., Grandori & Furnari, 2008).
Likewise, the concepts of neutral permutations and
causal asymmetry enrich understanding of the re-
lationship between configurational ideal types and
performance, again providing a closer look at the
causal processes involved. Finally, the theoretical
perspective proposed here speaks to a number of
substantive issues in both the management and
strategy literatures, such as ambidexterity and
planned organizational change, and I conclude by
discussing these implications as well as outlining a
way forward for empirical research.
RETHINKING CAUSALITY IN
TYPOLOGICAL THEORIES
Having an accurate understanding of causal rela-
tionships is essential for both strategic management
and organization theory (Durand & Vaara, 2009).
Viable competitive strategies rest on decision mak-
ers’ understanding of the appropriate causal rela-
tionships between those variables that management
controls, such as strategy and organizational struc-
ture, and those that are mostly outside the direct
control of management, such as the nature of an
industry (e.g., Galbraith & Schendel, 1983; Porter,
1980). Furthermore, these causal understandings
have to relate to the strategies of rivals and the
dimensions on which they compete, as well as to
those of decision makers’ own firms and their in-
ternal processes (Porac & Thomas, 1990; Porac,
Fiss 395
Thomas, & Baden-Fuller, 1989). Because of the im-
portance of causal processes, research from a vari-
ety of areas has focused on cause-effect relation-
ships as well as their understanding by strategic
decision makers (e.g., Barr et al., 1992; Durand &
Vaara, 2009; Grandori & Furnari, 2008; Huff, 1990;
King & Zeithaml, 2001; Reger & Huff, 1993).
Perhaps the most widely applied tool for map-
ping the competitive landscape and guiding strate-
gic analysis is the use of typologies and strategy
archetypes (e.g., Burns & Stalker, 1961; Hofer &
Schendel, 1978; Miles & Snow, 1978; Mintzberg,
1983; Porter, 1980). Furthermore, typologies are
also very popular among organizational researchers
and educators and have inspired a considerable
amount of both conceptual development and em-
pirical research (e.g., Ketchen, Thomas, & Snow,
1993). Here, I follow Doty and Glick’s definition of
typologies as “conceptually derived interrelated
sets of ideal types” [that] “identify multiple ideal
types, each of which represents a unique combina-
tion of the organizational attributes that are be-
lieved to determine the relevant outcome(s)” (1994:
232). This definition is helpful for several reasons.
First, it identifies typologies as complex theoretical
statements that emerge from a unique form of the-
ory building; as McKinney noted, “Types function
as theory” (1969: 8), a view that also informs my
understanding of typologies as ways to organize
complex cause-effect statements. Second, as Doty
and Glick noted, “The organizational types identi-
fied in typologies are developed with respect to a
specified organizational outcome” (1994: 232).
This insight likewise resonates with the conceptual
and empirical approach developed here, which fo-
cuses on configurations of causes in relation to an
outcome of interest. Finally, the definition clearly
distinguishes theoretically derived typologies from
empirically derived taxonomies or classification
systems, thus avoiding confusion between these
different concepts (McKelvey, 1975).1
The process of constructing a typology usually
involves the definition of an n-dimensional prop-
erty space from which the typology can be empiri-
cally reproduced. Since the purpose of a typology
is to reduce the complexity of the empirical world,
typification usually involves the pragmatic reduc-
tion of an extensive set of attributes to a limited
1
A full review of the literature on types and typology
is beyond the scope of this manuscript. However, a few
relevant sources include Bailey (1973), Bendix (1963),
Capecchi (1966), Doty and Glick (1994), Lazarsfeld
(1937), McKelvey (1982), McKinney (1966, 1969), Rich
(1992), Rose (1950), Weber (1949), and Winch (1947).
396 Academy of Management Journal
set relevant to the purpose at hand (Bailey, 1973;
McKinney, 1969). This process of reduction creates
two kinds of typologies: “monothetic” typologies,
in which each feature is necessary for membership
and the set of features is sufficient, and “poly-
thetic” typologies, which can be formed from dif-
ferent combinations of values on the features of
interest. Because they allow the grouping of cases
that are similar though perhaps not identical in
terms of their features, polythetic typologies tend to
ensure greater parsimony and are considered supe-
rior for research actually intended to identify spec-
imens as part of a type (Bailey, 1973). In the fol-
lowing, I focus on such polythetic typologies.
Typologies are theoretically attractive because
they move thought beyond traditional linear or
interaction models of causality such as contingency
theories (Doty & Glick, 1994). They can accommo-
date multiple between-construct relationships that
can be marked by complementary, additive, substi-
tution, or suppression effects, thus accommodating
considerable levels of causal complexity. As short-
hand devices, they furthermore allow managers
and researchers to cognitively simplify a complex
environment by highlighting commonalities be-
tween firms and allowing comparisons (e.g., Ham-
brick, 1983).
Because they are based on complex, synergistic
patterns of relationships, typologies orient one to-
ward a holistic understanding and ideal profiles
(Doty & Glick, 1994; McKelvey, 1982). Yet this ho-
lism is not without problems. For instance, in de-
scribing the difficulties of developing typological
theory, Doty and Glick noted that “as the number of
descriptive dimensions is increased, it becomes
more difficult to ensure that only those dimensions
that are causally related to the dependent variable
are included in the typology” (1994: 245). Simi-
larly, Scott (1981) argued that including dimen-
sions that are only spuriously correlated with the
outcome in question may lead to a misunderstand-
ing of the true causal processes involved. Summa-
rizing the key challenge of typological theories,
Doty and Glick argued that “the intuitive simplicity
of typologies masks some important complexities”
(1994: 245). Typologies can be deceptive.
As tools for understanding and summarizing
complex causal effects, typologies may lead schol-
ars astray for several reasons. To further examine
these, I draw on the causal attribution (e.g., Gopnik
& Schulz, 2007; Kim, Dirks, Cooper, & Ferrin, 2006;
Sloman, 2005; Waldman, Hagmayer, & Blaisdell,
2006) and the cognitive strategy literatures (e.g.,
Abrahamson & Fombrun, 1994; Barr et al., 1992;
Huff, 1990; Reger & Huff, 1993) for insight into the
construction of mental models of causal relation-
April
ships. Although this cognitive literature has been
largely taxonomic (e.g., Porac & Thomas, 1990) and
is conceptually different from typologies in that it
has focused on eliciting causal understandings and
“theories in use” from respondents, I suggest that
its insights regarding biases and tendencies to per-
ceive causal relationships are not restricted to clas-
sification but frequently also pervade the construc-
tion of typologies. Accordingly, I draw upon this
perspective to better understand why the construc-
tion of typologies may at times be problematic.
For instance, cognitive research suggests that
causal inferences may lead to illusory causation
when the true nature of causal relationships is
poorly understood or inappropriately defined (e.g.,
Nadkarni & Narayanan, 2007). Two mechanisms in
particular are relevant in this regard. First, when
applied to complex relationships, typologies may
fulfill the cognitive need for simplification when
one faces “information overload” (e.g., Schwenk,
1984). Second, decision makers with incomplete
information frequently engage in elaboration by—
often unconsciously—filling in gaps in the data
when interpreting stimuli (Reger & Huff, 1993; Ro-
sch, 1981). In combination, the two cognitive
mechanisms of simplification and elaboration are
likely to be present in typologies and may result in
causal understandings that are frequently inaccu-
rate and potentially misleading. Furthermore, be-
cause established notions of causal processes are
difficult to discard (Carley & Palmquist, 1992;
Hodgkinson, 1997), typologies are likely to pro-
mote cognitive inertia, a process that has been
shown by cognitive researchers to prevent decision
makers from acquiring new knowledge and explor-
ing alternatives (e.g., Reger & Palmer, 1996).
Although a holistic approach is thus a strength of
configurational theories, it can also inhibit the de-
velopment of accurate understandings of processes
because theorizing is more likely to end once a
configuration is identified, thereby limiting under-
standing as to what causal mechanisms are at work
and are driving effects (McPhee & Scott Poole,
2001; Reynolds, 1971). Specifically, “Most con-
figurational theories are what Althusser (1972)
called ‘expressive totalities’—they are supposed to
be consistent because each part reflects the under-
lying logic of the whole” (McPhee & Scott Poole,
2001: 515). However, a good theory should ques-
tion the assumption of consistency—that is, the
assumption that all parts of the configuration are
equally necessary or important.
These concerns regarding typological theories
are magnified by limited empirical support for the
causal processes involved. For instance, the con-
ceptual dimensions of many typologies are derived
2011
“without much empirical support beyond perhaps
some grounding in case studies and anecdotal ac-
counts of competitive activity” (Galbraith & Schen-
del, 1983: 155). At the same time, there is a lack of
theory allowing prediction and explanation of
which organizational practices are complementary
(Grandori & Furnari, 2008). Typologies are thus a
double-edged sword. At their best, they are memo-
rable, neat, and evocative (Miller, 1996). At their
worst, typologies are little more than simplistic
overviews that offer only a cursory look at organi-
zations (Rich, 1992).
The Core-Periphery Distinction
To develop a framework for understanding the
causal processes involved in typological theory, I
draw on the dual concepts of core and periphery.
The distinction between core and periphery ap-
pears particularly useful for understanding typolo-
gies for a number of reasons. At a most basic level,
cognitive researchers have argued that the human
mind’s ability to classify is better understood in
terms of a conceptual structure consisting of core
and peripheral categories (Hahn & Chater, 1997;
Hunn, 1982; Rosch, 1978, 1981). As such, the
distinction is grounded in a considerable body
of research on the cognitive foundations of
classification.
Furthermore, the notion of a core-periphery dis-
tinction has been successfully used in a number of
domains of the organization and strategy litera-
tures. For instance, classic arguments in organiza-
tion theory and design have used the notion of a
core and periphery in relation to an organization’s
primary technology and decision making (e.g., Pfef-
fer, 1976; Thompson, 1967), and current theorizing
has likewise emphasized the core-periphery dis-
tinction (e.g., Grandori & Furnari, 2008).
Perhaps the most influential view of core versus
periphery in organizations is that of Hannan and
Freeman (1984), who defined an organization’s
core as its mission, authority structure, technology,
and marketing strategy. In their definition, “Core-
ness means connectedness” (Hannan et al., 1996:
506), with change in core elements requiring ad-
justments in most other features of an organization.
This definition of core versus peripheral elements
has been adopted in a considerable number of sub-
sequent studies (e.g., Kelly & Amburgey, 1991;
Singh, House, & Tucker, 1986).
Likewise, strategy researchers have argued for
the distinction between core and peripheral con-
cepts in understanding strategic schemata (knowl-
edge structures) that top managements use in mak-
ing strategic decisions (e.g., Eden, Ackermann, &
Fiss 397
Cropper, 1992; Gustafson & Reger, 1995; Lyles &
Schwenk, 1992; Porac & Rosa, 1996). These re-
searchers have argued that depth and significance
are higher for core than for peripheral concepts
(Nadkarni & Narayanan, 2007a). Furthermore, prior
studies have suggested that the notion of core ver-
sus peripheral concepts is important with regard to
causal inferences (Carley & Palmquist, 1992) and
may draw a decision maker’s attention to nonexis-
tent relationships, because managers tend to auto-
matically infer new events by the use of core con-
cepts (Barr et al., 1992; Nadkarni & Narayanan,
2007a).
At a more aggregate level, strategy researchers
have pointed to the presence of core and peripheral
groupings in the structures of strategic groups (Dra-
nove, Peteraf, & Shanley, 1998; McNamara, Deep-
house, & Luce, 2003; Reger & Huff, 1993) and have
applied the core-periphery distinction to under-
standing diversification strategy in terms of over-
lapping product and industry segments (e.g., Sig-
gelkow, 2003; Wiersema & Liebeskind, 1995).
The concepts of core and periphery have also
figured prominently in the literature on social net-
works (e.g., Borgatti & Everett, 1999; Garcı́a Muñiz
& Ramos Carvajal, 2006) and in organizational net-
work analysis. For instance, Stuart and Podolny
(1996) employed the distinction to classify the
technology position of Japanese semiconductor
firms in terms of their knowledge domains, and
Gulati and Gargiulo (1999) identified core and pe-
riphery patterns in interorganizational alliances.
Similarly, the literature on the diffusion of innova-
tions has used a core-periphery model to explain
patterns of adoption of organizational practices
(e.g., Abrahamson & Rosenkopf, 1997; Galaskie-
wicz & Wasserman, 1989).
A Causal Core-Periphery Perspective
Core-periphery distinctions have thus been cru-
cial to efforts to understand a variety of substantive
issues in both organization theory and strategy, and
this distinction may also be useful in informing
typological theories. For instance, several studies
have suggested that firms vary in the degree to
which they identify or are associated with a strat-
egy: some firms follow the strategy closely and can
be considered “core” firms, but others follow the
strategy less closely and can be categorized as sec-
ondary or “peripheral” (McNamara et al., 2003; Pe-
teraf & Shanley, 1997). In particular, Reger and Huff
(1993) argued that group membership is a matter of
degree and that strategic groups are akin to “fuzzy
sets”—an argument that I likewise pursue here and
398 Academy of Management Journal
actually implement in the measurement of type
membership.
Similarly, Siggelkow (2002) argued that it is nec-
essary to develop a better understanding of the
nature of core elements in organizational configu-
rations. Drawing upon network measures and the
notion that “coreness means connectedness” (Han-
nan et al., 1996), Siggelkow defined an organiza-
tional core element as “an element that interacts
with many other current or future organizational
elements” (2002: 126 –127). Core elements of a con-
figuration are thus surrounded by a series of elab-
orating or peripheral elements that reinforce the
central features of the core (Grandori & Furnari,
2008).
Applying these insights to the literature on typol-
ogies, I argue that the notions of core and periphery
are useful for enhancing understanding of causality
in typologies. This argument extends prior theoriz-
ing on typologies, which has suggested that the
contribution of a specific attribute to an ideal type
needs to be weighted by its theoretical importance
(Doty & Glick, 1994), thus pointing toward a view
of typologies as being made of elements with dif-
fering significance. Furthermore, this literature
suggests that ideal types are “pure” forms of a con-
figuration (Miles & Snow, 2003: 30) and that devi-
ation from ideal types usually results in lower per-
formance (Doty et al., 1993; Van de Ven & Drazin,
1985).
Building on these insights regarding the differing
importance of configurational elements can also
extend understanding of causal relationships
within typologies. Specifically, I suggest here a def-
inition of coreness based on which elements are
causally connected to a specific outcome. In accor-
dance with this understanding, I define core ele-
ments as those causal conditions for which the
evidence indicates a strong causal relationship
with the outcome of interest. In contrast, peripheral
elements are those for which the evidence for a
causal relationship with the outcome is weaker.
Maintaining a concern for what makes elements
causally relevant, this understanding also impor-
tantly shifts the focus from connectedness with
other organizational elements to the causal role
they play in the configuration relative to the out-
come. It moreover fits closely with the central in-
sights of organization theory that core elements are
those that are the most important to organizational
performance and survival (Romanelli & Tushman,
1994) and that types are understood in relation to
an outcome (Doty & Glick, 1994). In sum, this
suggests:
April
Proposition 1. Typological configurations are
characterized by a core and a periphery.
Furthermore, in focusing on the causal relation-
ships among configurational elements and out-
comes, the theoretical perspective introduced here
has implications for understanding of equifinality
in typologies and configurations. A model of causal
core and periphery emphasizes the idea that sev-
eral causal paths to an outcome exist—that is, equi-
final configurations exist (e.g., Doty et al., 1993;
Gresov & Drazin, 1997; Payne, 2006). However, the
current perspective enriches the study of equifinal
configurations through the notion of neutral per-
mutations of a given configuration. Specifically, I
suggest that, within any given configuration, more
than one constellation of different peripheral
causes may surround the core causal condition,
and the permutations do not affect the overall per-
formance of the configuration. This argument
builds on prior work that has shown how trade-offs
between, for example, strategies and functions fre-
quently lead to several equifinal configurations that
each results in effective performance (e.g., Delery &
Doty, 1996; Marlin et al., 2007). As these research-
ers have argued, hybrid types, all of which can
result in the specified level of a dependent variable,
may mark typological configurations (Doty & Glick,
1994). This argument suggests a need for further
exploration of viable strategic alternatives, particu-
larly with a focus on understanding intratype sim-
ilarities and dissimilarities (Reger & Huff, 1993).
The notion of neutral permutations introduced
here builds on these arguments by enhancing the
current concept of equifinality in several ways.
First, the concept of neutral permutations provides
a finer-grained understanding of the different
cause-effect relationships related to an outcome by
distinguishing between first- and second-order
equifinality. By first-order equifinality, I mean
equifinal types that exhibit different core character-
istics (e.g., type A vs. type B). By second-order
equifinality, I mean neutral permutations within a
given first-order equifinal type (e.g., type A1 vs.
A2 . . . An).
Second, the concept of neutral permutations con-
tributes to the theory of equifinality by implying
that, although different permutations may be equi-
final regarding an outcome, they are not equifinal
regarding future states of development (Stadler,
Stadler, Wagner, & Fontana, 2001). As such, an
understanding of the causal nature of a configura-
tion is essential for understanding trajectories of
organizational change, an important issue that has
largely been neglected in the study of configura-
tions (Grandori & Furnari, 2008). In sum, my argu-
2011
ments suggest the following regarding the nature of
typological configurations:
Proposition 2. Core and periphery in typologi-
cal configurations frequently exhibit neutral
permutations.
The theoretical perspective on typologies intro-
duced here also has implications for how cause-
effect relationships combine to achieve outcomes.
Specifically, it builds upon a recent and growing
interest in understanding the causal necessity and
sufficiency of, rather than the correlations between,
configurations and performance (e.g., Fiss, 2007;
Kogut, MacDuffie, & Ragin, 2004). Such an argu-
ment is both attractive and important because it
implies causal asymmetry (Ragin, 2008)—that
causes leading to the presence of an outcome of
interest may be quite different from those leading to
the absence of the outcome. As noted above, in
contrast, a correlational understanding of causality
implies causal symmetry because correlations tend
to be symmetric. For instance, if one were to model
the inverse of high performance, then the results of
a correlational analysis would be unchanged, ex-
cept for the sign of the coefficients. However, a
causal understanding of necessary and sufficient
conditions is causally asymmetric—that is, the set
of causal conditions leading to the presence of the
outcome may frequently be different from the set of
conditions leading to the absence of the outcome.
For instance, even though the presence of a partic-
ular combination of causes may lead to high per-
formance, it may not be merely the absence of this
combination, but the presence of an entirely differ-
ent set of causes, that leads to low performance.
Introducing this notion of causal asymmetry builds
on prior work arguing that typological theories al-
low movement beyond traditional linear or interac-
tion theories by shifting understanding toward con-
figurational thinking and nonlinear relationships
among constructs (Doty et al., 1993; Meyer et al.,
1993). Such complex relationships and patterns
can frequently not be represented with traditional
bivariate contingency theories, as two organization-
al characteristics can be positively related in one
ideal type, negatively related in another, and unre-
lated in a third (e.g., Delery & Doty, 1996; Doty &
Glick, 1994). Thus, in a configurational approach
the causal relationships are viewed not so much in
terms of correlations as in terms of sets of equally
effective patterns (Doty et al., 1993; Van de Ven &
Drazin, 1985). The concept of causal asymmetry
builds on this understanding by indicating that
equifinality may change depending on outcome
levels; as one moves across outcome levels, differ-
Fiss 399
ent sets of equally effective configurations may
arise.
Shifting to a causal core-periphery view of organ-
izational typological configurations thus allows for
such differing sets of causal conditions; one set
may lead, for instance, to average performance,
while a different set may lead to high performance,
and yet another set may lead to very high perfor-
mance. These arguments suggest:
Proposition 3. Typological configurations are
marked by causal asymmetry.
THE MILES AND SNOW TYPOLOGY OF
ORGANIZATIONAL CONFIGURATIONS
To empirically test the theoretical perspective I
develop here, I draw on the Miles and Snow (1978,
2003) typology of generic organizational configura-
tions. This is perhaps the most widely used typol-
ogy of organizations, and classic studies on config-
urations have tested it and found considerable
support for it (e.g., Doty et al., 1993; Hambrick,
1983; Ketchen et al., 1993). In fact, as Hambrick
(2003) noted, it presents one of the most widely
tested, validated, and enduring strategy frame-
works of the last 25 years; researchers have found
strong and consistent support for the typology in
settings ranging from hospitals to industrial prod-
uct and life insurance companies. Moreover, sev-
eral recent studies have given the Miles and Snow
typology renewed attention, using it to generate
new knowledge (e.g., DeSarbo et al., 2005; Hult,
Ketchen, Cavusgil, & Calantone, 2006; Kabanoff &
Brown, 2008; Olson, Slater, & Hult, 2005; Slater &
Olson, 2000). In sum, given its wide usage and
continued relevance, the Miles and Snow typology
is particularly suitable as a context in which to
apply the theoretical perspective presented here.
Miles and Snow’s typology is based on three
organizational types: “Prospector,” “Analyzer,”
and “Defender.” A fourth, “Reactor,” is largely a
residual type because, in contrast to the other three,
the Reactor “lacks a consistent strategy-structure
relationship” (Miles & Snow, 2003: 29) and thus pres-
ents an absence of strategy rather than a viable strat-
egy (Inkpen & Choudhury, 1995; Zajac & Shortell,
1989). Table 1 provides an overview of the three ideal
profiles as they relate to structure and strategy.
A Prospector is typically a small but growing
organization continually in search of new product
and market opportunities. Change is a prime chal-
lenge for this organization, and the administrative
challenge is thus how to facilitate operations rather
than how to control them. Accordingly, Prospec-
tors tend to score relatively low with regard to
400 Academy of Management Journal
formalization and centralization. The need to alter
organizational structure in response to a changing
environment means that they tend to have little
division of labor and flat organizational structures,
resulting in low organizational complexity. In
terms of strategy, their focus on innovation and
product features makes them highly similar to Por-
ter’s differentiators, and the same features mean
they score on a cost leadership strategy (e.g., Miller,
1986; Parnell, 1997; Segev, 1989).
In contrast, a Defender is more typically a large
and established firm that aims to protect its prom-
inence in a product-market. Prospectors are fo-
cused on change, Defenders on stability, which is
reflected in their organizational structures. Man-
agement usually aims for highly centralized control
of organizational operations and more often uses
formalized processes and policies to specify appro-
priate behaviors for organization members. Defend-
ers are usually marked by an extensive division of
labor and numerous hierarchies, indicating high
administrative complexity. In terms of strategy, De-
fenders typically pursue cost leadership rather than
differentiation (e.g., Miller, 1986; Segev, 1989;
Shortell & Zajac, 1990).
As Miles and Snow noted, Prospectors and De-
fenders “reside at opposite ends of a continuum” of
strategies; Analyzers, however, lie “between these
two extremes” (2003: 68). Table 1 reflects this struc-
ture in showing “Analyzer” in-between “Prospector”
and “Defender” in terms of strategic and structural
attributes, except for complexity. The reason for this
lies in the hybrid nature of the Analyzer, which has to
be able to accommodate both stability and change,
indicating an ideal profile marked by rather complex
structures (Miles & Snow, 2003: 79).
Open Questions Regarding the Miles and
Snow Typology
Although the overall Miles and Snow typology
has found widespread usage, the evidence on its
TABLE 1
Ideal Profiles Based on Miles and Snow
Characteristic Prospector Analyzer Defender
Structure
Size Low Medium High
Formalization Low Medium High
Centralization Low Medium High
Complexity Low High High
Strategy
Differentiation High Medium Low
Low cost Low Medium High
April
specific propositions is in fact less than clear (e.g.,
Hambrick, 1983; Zahra & Pearce, 1990). In review-
ing the literature and empirical studies on the ty-
pology, Zahra and Pearce (1990) noted that there
has been essentially no research on how firms of
different strategic types utilize different organiza-
tional structures and coordination mechanisms. As
such, the important question of which elements of
the typology are relevant and which elements are
either peripheral or even irrelevant has received
essentially no attention. This problem is further
magnified by the widespread use of a self-typing
instrument wherein survey respondents identify
themselves with strategic archetypes (Zahra &
Pearce, 1990), which largely rules out gaining a
better understanding of the causal processes in-
volved in the different types.
Furthermore, according to the typology the three
strategy types have equal effectiveness. However,
Hambrick (1983) found that Analyzers tended to
outperform both Prospectors and Defenders on per-
formance measures such as return on investment
and market share and suggested that “in general the
‘superior’ strategy was neither of the two extreme
strategies” (Hambrick, 1983: 18). Similarly, Ka-
banoff and Brown (2008) found that Analyzers per-
formed relatively well in profitability when com-
pared with the other types, as did Snow and
Hrebiniak (1980), although their sample for this
strategic type was relatively small. These studies
suggest that taking a middle position that “com-
bines the strengths of both the Prospector and the
Defender into a single system” (Miles & Snow,
2003: 68) results in higher performance than taking
either extreme position, thus supporting the argu-
ments of the ambidexterity literature, which points
to the possibility of achieving superior perfor-
mance by simultaneously achieving efficiency and
adaptiveness (e.g., Raisch & Birkinshaw, 2008;
Tushman & O’Reilly, 1996).
In contrast, other authors have pointed to the
importance of trade-offs in strategy. For instance,
Porter (1980, 1996) argued that differentiation and
cost leadership can be combined only rarely. As a
result, organizations should pursue either a differ-
entiation or cost-leadership strategy but should not
try to combine both strategies so as to avoid getting
“stuck in the middle” with lower performance.
Similarly, March (1991) argued for a fundamental
trade-off between the “exploration” and “exploita-
tion” strategies. As DeSarbo et al. (2005) pointed
out, more research is thus needed on the topic of
strategic type and performance and how different
elements of the three strategic types relate to one
another.
2011
Yet another important but unresolved aspect of
the Miles and Snow typology relates to whether it
is universally applicable across environments or is
context dependent. Hambrick (1983: 7) noted that
the generic character of the typology ignores indus-
try and environmental peculiarities, and Zajac and
Shortell similarly pointed out that Miles and
Snow’s notion of generic strategies tends to “as-
sume that the various strategies are equally viable
across environmental contexts and, by implication,
across time” (1989: 413). However, as DeSarbo et al.
(2005) pointed out, very few studies have empiri-
cally examined this relationship between the na-
ture of an environment and strategic type. In fact,
apart from the study by Doty et al. (1993), no other
research appears to have simultaneously examined
configurations of organizational structure, strategy,
and environment. Yet a configurational approach
as posited in the typology demands a detailed un-
derstanding of the causal relationships among in-
ternal organizational features, such as structural
characteristics and strategy type, on the one hand,
and external, environmental characteristics on the
other. In sum, these open questions indicate that,
despite its wide use and continued influence, the
Miles and Snow typology faces a number of chal-
lenges characteristic of typological theories, thus
making it a suitable choice for the current study.
Modeling Causal Configurations
Answers to the questions outlined here are to a
considerable extent influenced by the analytical
approach employed. Apart from self-typing, two
main approaches have been used to study the Miles
and Snow typology and its relationship to perfor-
mance. The first is inductive and primarily uses
cluster analysis to derive an empirical solution
(e.g., Ketchen et al., 1993). The second approach is
deductive and uses deviation score analysis to ex-
amine fit with a theoretically defined profile (e.g.,
Doty et al., 1993). Although both approaches have
considerably enhanced understanding of typologi-
cal and configurational theory, they nevertheless
also have limited ability to provide insights into the
causal nature of a configuration—that is, they are
not well suited to shedding light on just which
aspect of a configuration leads to high performance
(Fiss, 2007, 2009).
In the current article, I complement the theoret-
ical perspective I introduce by using set-theoretic
methods for studying cases as configurations. The
current study thus builds on the set-theoretic meth-
ods introduced by Ragin (1987) and later extended
(Ragin, 2000, 2008; Ragin & Fiss, 2008; Rihoux &
Ragin, 2009). Set-theoretic methods such as fuzzy
Fiss 401
set QCA are uniquely suitable for testing typologi-
cal and configurational theory because they explic-
itly conceptualize cases as combinations of attri-
butes and emphasize that it is these very
combinations that give cases their uniqueness. Set-
theoretic methods thereby differ from conven-
tional, variable-based approaches in that they do
not disaggregate cases into independent, analyti-
cally separate aspects but instead treat configura-
tions as different types of cases. These features
make set-theoretic methods attractive for organiza-
tional and strategy researchers, as several recent
studies applying QCA and fuzzy sets in organiza-
tional settings have demonstrated (e.g., Bakker,
Cambré, Korlaar, & Raab, 2010; Crilly, 2011; Fiss,
2007, 2009; Grandori & Furnari, 2008; Greckhamer,
2011; Greckhamer, Misangyi, Elms, & Lacey, 2008;
Jackson, 2005; Kogut et al., 2004; Marx, 2008; Pa-
junen, 2008; Schneider, Schulze-Bentrop, & Pau-
nescu, 2009). The methodological approach used
here thus sheds new light on the causal relation-
ship between the characteristics of a configuration
and an outcome of interest.
DATA AND METHODS
I used data on a sample of 205 high-technology
manufacturing firms located in the United King-
dom. The data were collected in 1999 using a sur-
vey sent to the CEOs and managing directors of
these firms (Cosh, Hughes, Gregory, & Jayanthi,
2002) and were especially useful for my purposes
for several reasons. First, the data contain a rich set
of measures on the firms’ strategy, structure, envi-
ronment, and performance, thus allowing me to
examine the effectiveness of different configura-
tions. Second, the data are restricted to manufac-
turing firms, assuring comparability regarding op-
erations by excluding, for example, service firms,
which frequently have very different operational
requirements. Finally, although the data come from
the high-technology manufacturing sector, they in-
clude firms operating in several industries, thus
offering variation in the rate of change and uncer-
tainty of the firms’ competitive environments that
would not be available in a single-industry study.
Although the data are uniquely appropriate for
the current study, they also have some limitations.
The survey’s response rate of 14 percent was some-
what lower than is usually desirable, although it is
still slightly above the 10 –12 percent response rate
that is typical for surveys mailed to CEOs in the
United States (e.g., Geletkanycz, 1997; Hambrick,
Geletkanycz, & Fred-Levinson, 1993). For reasons
of confidentiality, the original investigators were
not able to provide response bias analyses or firm
402 Academy of Management Journal
names, and I could therefore not conduct my own
analyses. However, the representativeness of the sam-
ple is less of a threat to validity in the current study
than it would be usually for at least three reasons:
First, my interest is not the U.K. high-technology
sector per se, but the sector as a setting in which to
test arguments relating to configurational theory.
Accordingly, the presence of some response bias
(e.g., a greater number of small firms or firms with
high performance responded) would not threaten
the validity of findings. Nevertheless, comparisons
with available aggregate statistics indicate that the
sample, which included firms with considerable
variation in size, structure, and environment, is
representative of the underlying population in
terms of organization size. Second, some of the
most influential and path-breaking studies, such as
those of Ketchen et al. (1993) and Doty et al. (1993),
have in fact used nonrandom samples of organiza-
tions selected on the basis of geographical proxim-
ity and social contacts, indicating that use of ran-
dom samples is not an essential feature of the
current research context. Third, and perhaps most
importantly, in contrast to standard econometric
methods, such as regression analysis, the nonpara-
metric, fuzzy set methods I employ here make sam-
ple representativeness less of an issue. This is the
case because— unlike, for example, regression anal-
ysis—fuzzy set QCA does not rest on an assump-
tion that data are drawn from a given probability
distribution. Furthermore, as I explain below, I em-
ployed calibrated sets to measure my constructs of
interest. This calibration reduces sample depen-
dence, as set membership is defined relative to
substantive knowledge rather than the sample
mean, thus further reducing the importance of sam-
ple representativeness. In sum, these points suggest
that the advantages of the data heavily outweigh
their limitations.
Analysis
The current study employs a set-theoretic ap-
proach based on fuzzy set QCA, an analytic tech-
nique grounded in set theory that allows for a de-
tailed analysis of how causal conditions contribute
to an outcome in question. This approach is
uniquely suited for analyzing causal processes in
typologies because it is based on a configurational
understanding of how causes combine to bring
about outcomes and because it can handle signifi-
cant levels of causal complexity (e.g., Fiss, 2007;
Ragin, 2000, 2008). The basic intuition underlying
QCA is that cases are best understood as configu-
rations of attributes resembling overall types and
that a comparison of cases can allow a researcher to
April
strip away attributes that are unrelated to the out-
come in question. In its logic, this approach is
based on the “method of difference” and the
“method of agreement” outlined by John Stuart
Mill (1843/2002), in which one examines instances
of the cause and outcome to understand patterns of
causation.2 However, set-theoretic analysis exam-
ines causal patterns by focusing on the set-subset
relationship. For instance, to explain what config-
urations lead to high performance, it examines
members of the set of “high-performing” organiza-
tions and then identifies the combinations of attri-
butes associated with the outcome of interest (high
performance) using Boolean algebra and algorithms
that allow logical reduction of numerous, complex
causal conditions into a reduced set of configura-
tions that lead to the outcome.
To empirically accomplish this identification of
causal processes, QCA proceeds in three steps. Af-
ter the independent and dependent measures have
been transformed into sets as described above, the
first step is using these set measures to construct a
data matrix known as a truth table with 2k rows,
where k is the number of causal conditions used in
the analysis. Each row of this table is associated
with a specific combination of attributes, and the
full table thus lists all possible combinations. The
empirical cases are sorted into the rows of this truth
table on the basis of their values on these attributes,
with some rows containing many cases, some rows
just a few, and some rows containing no cases if
there is no empirical instance of the particular com-
bination of attributes associated with a given row.
In a second step, the number of rows is reduced
in line with two conditions: (1) the minimum num-
ber of cases required for a solution to be considered
and (2) the minimum consistency level of a solu-
tion. “Consistency” here refers to the degree to
which cases correspond to the set-theoretic rela-
tionships expressed in a solution. A simple way to
estimate consistency when using fuzzy sets is as
the proportion of cases consistent with the out-
come—that is, the number of cases that exhibit a
given configuration of attributes as well as the out-
2
For further background on QCA, the reader is re-
ferred to Ragin (2000, 2008). For a shorter introduction as
well as tutorials and empirical examples, the reader is
referred to Crilly (2011), Fiss (2007), Greckhamer (2011),
Greckhamer et al. (2008), Herrmann and Cronqvist
(2009), Jackson (2005), Rihoux and Ragin (2009), and
Schneider and Wagemann (2007). Although QCA was
initially conceived as a small-N approach (e.g., between
15 and 40 cases), the current study follows more recent
works that have extended QCA to large-N settings (e.g.,
Ragin & Fiss, 2008).
2011
come divided by the number of cases that exhibit
the same configuration of attributes but do not ex-
hibit the outcome. The current study uses a refined
measure of consistency introduced by Ragin (2006)
that gives small penalties for minor inconsistencies
and large penalties for major inconsistencies. I set
the lowest acceptable consistency for solutions at
ⱖ0.80, which is again above the minimum recom-
mended threshold of 0.75 (e.g., Ragin, 2006, 2008).
Also, the minimum acceptable solution frequency
was set at three. Overall, 56 cases fell into configura-
tions exceeding the minimum solution frequency. Of
these cases, 40 also exceeded the minimum consis-
tency threshold of 0.80 for higher performance, and
33 cases exceeded this threshold for very high perfor-
mance. No case exceeded the consistency threshold
for the absence of high or very high performance, a
finding I further discuss in the results section.
In a third step, an algorithm based on Boolean
algebra is used to logically reduce the truth table
rows to simplified combinations. The current
study uses the truth table algorithm described by
Ragin (2005, 2008). This algorithm is based on a
counterfactual analysis of causal conditions,
which has the advantage of allowing for a cate-
gorization of causal conditions into core and pe-
ripheral causes. Counterfactual analysis is rele-
vant to configurational analysis because even
relatively few elements of a configuration quickly
lead to an astronomically large number of truth
table rows. For researchers, this characteristic
means that there will frequently be very few or no
empirical instances of any particular configura-
tion. This challenge of configurational ap-
proaches is known as the “problem of limited
diversity” (e.g., Ragin, 2000), and counterfactual
analysis offers a way to overcome the limitations
of a lack of empirical instances.
To deal with the problem of limited diversity
using counterfactual analysis, the truth table algo-
rithm distinguishes between parsimonious and in-
termediate solutions on the basis of “easy” and
“difficult” counterfactuals (Ragin, 2008). “Easy”
counterfactuals refer to situations in which a re-
dundant causal condition is added to a set of causal
conditions that by themselves already lead to the
outcome in question. As an example, assume one
has evidence that the combination of conditions
A•B•⬃C (read: A and B but not C) leads to the
presence of the outcome. No evidence exists as to
whether the combination A•B•C (read: A and B
and C) would also lead to the outcome, but theo-
retical or substantive knowledge links the presence
(not the absence) of C to the outcome. In such a
situation, an easy counterfactual analysis indicates
that both A•B•⬃C and A•B•C will lead to the
Fiss 403
outcome, and the expression can be reduced to
A•B, because whether C is absent or present has no
effect on the outcome. In easy counterfactual anal-
ysis, the researcher thus asks, Would adding an-
other causal condition make a difference? If the
answer is no, one can proceed with the simplified
expression.
In contrast, “difficult” counterfactuals refer to
situations in which a condition is removed from a
set of causal conditions leading to an outcome on
the assumption that this condition is redundant.
For instance, one might have evidence that the
combination A•B•C leads to the outcome, but no
evidence as to whether the combination A•B•⬃C
also does so. This case is of course the inverse of
the situation above. In a difficult counterfactual
analysis, a researcher asks, Would removing a
causal condition make a difference? This question
is more difficult to answer. Theoretical or substan-
tive knowledge links the presence, not the absence,
of C to the outcome, and lacking an empirical in-
stance of A•B•⬃C, it is much harder to determine
whether C is in fact a redundant condition that can
be dropped, thus simplifying the solution to merely
A•B.
Distinguishing between easy and difficult coun-
terfactuals allows establishment of two kinds of
solutions. The first is a parsimonious solution that
includes all simplifying assumptions regardless of
whether they are based on easy or difficult coun-
terfactuals. The second is an intermediate solution
that only includes simplifying assumptions based
on easy counterfactuals.3 The notion of causal con-
ditions belonging to core or peripheral configura-
tions is based on these parsimonious and interme-
diate solutions: core conditions are those that are
part of both parsimonious and intermediate solu-
tions, and peripheral conditions are those that are
eliminated in the parsimonious solution and thus
only appear in the intermediate solution. Accord-
ingly, this approach defines causal coreness in
terms of the strength of the evidence relative to the
outcome, not connectedness to other configuration-
al elements.
Outcome Measures
The primary outcome of interest in my study is
organizational performance, measured as return on
3
A third solution, of course, is the most complex one
that includes neither easy nor difficult counterfactuals.
However, such a solution is usually needlessly com-
plex and provides rather little insight into causal
configurations.
404 Academy of Management Journal
assets (ROA) and calculated as pretax profits
(losses) before deduction of interest and directors’
emoluments divided by total assets.4 I calibrated
this measure by “benchmarking” it to the overall
performance of the high-technology manufacturing
sector rather than by using a sample-dependent
anchor such as the mean for firms in the sample.
Data on average industry performance in the U.K.
high-technology manufacturing sector came from
ICC Business Ratio reports. I averaged ROA across
the main manufacturing industries in this sector,
such as computer equipment, printed circuits, sci-
entific and electronic instruments, electronic com-
ponents, and aerospace materials. The average
ROA for the sector was 7.8 percent, which is very
similar to a median ROA of 7.2 percent for the U.S.
high-technology sector in the same time period.
Because data regarding variation in performance
were not available for the U.K. sector, I used upper-
and lower-quartile information for the U.S. sector,
obtained from RMA Annual Statement Studies.
The analysis with fuzzy set QCA requires trans-
forming variables into sets calibrated regarding
three substantively meaningful thresholds: full
membership, full nonmembership, and the cross-
over point—that is, “the point of maximum ambi-
guity (i.e., fuzziness) in the assessment of whether
a case is more in or out of a set” (Ragin, 2008: 30).
The crossover point thus qualitatively anchors a
fuzzy set’s midpoint between full membership and
full nonmembership (see Ragin, 2000: 158). Fol-
lowing this approach, I created two fuzzy set mea-
sures of above-average firm performance. The first,
membership in the set of firms with high perfor-
mance was coded 0 if a firm showed average or
below-average performance (ROA ⱕ 7.8%; i.e.,
about the 50th percentile) and was coded 1 if the
4 by memos, and whether reporting relationships are
Because CEOs and managing directors self-reported
firm performance data, common method bias was possi-
ble. To assess this possibility, I used Harman’s single-
factor test (e.g., Hult et al., 2006; Konrad & Linnehan,
1995). If common method bias is a serious problem, a
single latent factor should account for a large proportion
of the variance in the sample. However, the unrotated
principal component factor solution suggests that such
bias is not an issue. The analysis of the full sample and
all measures results in four factors with eigenvalues
larger than 1, rather than a single factor. Furthermore, the
largest factor accounted for 22.4 percent of the variance,
and the largest three factors together accounted for 51
percent of the variance, indicating again that there is not
one general factor. Although the results do not perfectly
rule out the existence of common method bias, they do
suggest that it is unlikely to affect the results in any
substantive way.
April
firm showed high performance (ROA ⱖ 16.3%; i.e.,
the 75th percentile or higher). As the crossover
point, I chose the halfway mark of about 12 percent.
The second set measure, membership in the set
with very high performance, was again coded 0 for
average or below-average performance (ROA ⱕ
7.8%; i.e., about the 50th percentile) and 1 for an
ROA of 25 percent—arguably very high perfor-
mance, even though I was not able to obtain data
regarding the corresponding sector percentile. As
the crossover point for very high performance, I
chose an ROA of 16.3 percent (i.e., the 75th percen-
tile, or full membership in the previous set of high-
performing firms).
To additionally examine what causes led to the
absence of high performance, I also created mea-
sures of membership in the sets of firms with not-
high performance and low performance. Not-high
performance is simply coded as the negation of the
measure of high performance described above (1 for
average or below-average performance and 0 for
high performance). I therefore also created a mea-
sure of low performance (ROA of 0.0% ⫽ full mem-
bership, ROA of 7.8% ⫽ full nonmembership,
crossover set at 3.9%). In sum, the four outcome
measures cover a full spectrum of performance
outcomes.
Independent Measures
I assessed organizational structure using four
variables usually employed with the Miles and
Snow typology as well as other classic studies of
organizational structure (e.g., Pugh, Hickson, &
Hinings, 1968). The first one, formalization, was
measured using nine survey questions on the ex-
tent to which a firm uses, for example, formal pol-
icies and procedures to guide decisions and to de-
termine how far communications are documented
formally defined and plans are formal and written.
The answer to each question was rated on a scale
anchored by 1, “almost never”; 3,“about half the
time”; and 5, “nearly always.” I combined the nine
survey questions into a scale that showed very good
reliability (␣ ⫽ .83). Drawing on the scale, I created
a measure of membership in the set of firms with
high formalization, coding membership as fully
out for a response of “almost never” and fully in
for a response of “nearly always.” The crossover
point was the middle of the scale (“about half the
time”).
The second measure, centralization, was based
on five survey questions about the last decision
maker whose permission must be obtained for or-
ganizational decisions such as the addition of a
2011
new product or service, unbudgeted expenses, and
the selection of the type or brand of new equip-
ment. Answers to these questions were again mea-
sured on a five-point scale, but actual responses
were essentially restricted to four levels (depart-
ment head, division head, CEO, and board of direc-
tors). The items were again combined into a scale
that showed acceptable reliability with a Cron-
bach’s coefficient alpha of .74, which is above the
frequently recommended value of .70 (e.g., Nun-
nally, 1978). Using this scale, I coded firms with
decision making at the level of the department head
as fully out of the set of highly centralized firms
and firms with decision making at the board level
as fully in the set, taking the scale midpoint (be-
tween division head and CEO) as the crossover
point.
The third measure, administrative complexity,
was the product of vertical and horizontal differen-
tiation (Singh, 1986; Wong & Birnbaum-More,
1994). Following Pugh et al. (1968), I measured
vertical differentiation as the number of levels in
the longest line between direct worker and CEO,
and horizontal differentiation as the number of
functions with at least one full-time employee. For
the fuzzy set of firms with a high degree of admin-
istrative complexity, firms in the 1st percentile
(one level /1 function) were fully out, and firms in
the 99th percentile (six levels/17 functions) were
fully in. As a crossover point, I chose the product of
the 50th percentile values of each of the individual
measures (three levels times 9 functions—i.e., a
score of 27 on the complexity measure), which is
largely consistent with the mean score of prior stud-
ies using this complexity measure (e.g., Wong & Birn-
baum-More, 1994). This score based on the medians
of the two individual measures is also very close to
the median of the combined complexity measure
(which was 33), and robustness checks indicated that
results do not depend on the choice of coding.
The last measure of organizational structure was
size, which was based on a firm’s average number
of full-time employees, with groupings based on
the European Union enterprise size classes (1–9,
10 – 49, 50 –249, 250⫹). Specifically, firms with 250
or more employees were coded as fully in the set of
large firms, and those with fewer than 10 employ-
ees were coded as fully out; the midpoint was set at
50 employees.
Regarding my measures of firm strategy, I used
Porter’s (1980) strategy framework, which is con-
sistent with Miles and Snow’s typology and prior
research (David, Huang, Pei, & Reneau, 2002;
Miller, 1988). For instance, Doty et al. stated that
“Defenders compete by producing low-cost goods
or services and obtain efficiencies by relying on
Fiss 405
routine technology and economies of scale gained
from largeness” (1993: 1226). Likewise, Shortell
and Zajac noted that a Defender “emphasizes tight
controls and continually looks for operating effi-
ciencies to lower costs,” in contrast to a Prospector,
which “frequently adds to and changes its products
and services, consistently attempting to be first in
the market” (1990: 818). Furthermore, in a system-
atic comparison of the Porter and the Miles and
Snow typologies, Segev (1989) suggested a typol-
ogy that equates Defenders with cost leadership/
cost focus and Prospectors with differentiation/dif-
ferentiation focus.5
The measures for Porter’s two generic strate-
gies, cost leadership and differentiation, were
based on a factor analysis of six items relating to
the firms’ competitive capabilities. The first four
are related to low labor cost, low material con-
sumption, low energy consumption, and low in-
ventory cost, and the last two are related to new
product introduction and product features. A
principal component factor analysis with vari-
max rotation showed a two-factor solution, with
all items loading highly and cleanly on the two
factors, as shown in Table 2.
The items were combined into two scales that
showed strong reliability (cost leadership: ␣ ⫽ .86;
differentiation, ␣ ⫽ .80). Using these scales, I cre-
ated two fuzzy set measures. Membership in the set
of firms with a cost leadership strategy was coded
as fully out for a value of 1 (“not important”) and
fully in for a value of 5 (“critically important”); the
scale midpoint of 3 was the crossover point. Coding
of membership in the set of firms with a differen-
tiation strategy followed the same approach.
I measured environmental context using the two
constructs of environmental rate of change and un-
certainty, which in combination assess the dyna-
mism of a competitive environment (Baum &
Wally, 2003; Dess & Beard, 1984). By making
choices regarding both aspects of their product-
market domains, the firms in my sample largely
committed to either a stable or continuously chang-
ing domain, a key aspect of the entrepreneurial
challenge all firms face (Miles & Snow, 2003). As
prior studies have shown, the high-technology sec-
5
A slightly different view of the relationship between
the Miles and Snow typology and Porter’s strategies has
been proposed by Walker and Ruekert (1987), who de-
veloped a hybrid model that subdivides the defender
type into low-cost defenders and differentiated defend-
ers. Prior empirical support for this distinction has been
limited (e.g., Olson, Slater, & Hult, 2005; Slater & Olson,
2000), yet I examine the implications of this view in my
discussion of the results below.
406 Academy of Management Journal April

TABLE 2 predictability of technological change for manufac-

Principal Component Factor Analysis for the
Strategy Constructa
Survey Item Factor 1 Factor 2
1. Low labor cost 0.81 ⫺0.08
2. Low materials consumption 0.84 0.02
3. Low energy consumption 0.86 ⫺0.03
4. Low inventory cost 0.84 ⫺0.01
5. New product introduction ⫺0.07 0.91
6. Product features 0.03 0.91
Eigenvalue 2.81 1.67
Proportion of variance explained 0.47 0.28
by eigenvector
a
All items were measured on a scale ranging from 1 (“not
important”) to 5 (“critically important”).
tor shows considerable variation with regard to
both constructs (e.g., Fine, 1998; Mendelson & Pil-
lai, 1999). Rate of change measures the speed of
product and competitive change (Bourgeois &
Eisenhardt, 1988; Jurkovich, 1974).6 In environ-
ments with a high rate of change, firms face diffi-
culties earning above-average profits for prolonged
periods via a single innovation or product (Bogner
& Barr, 2000). I operationalized the rate of change
as the length of a firm’s main product life cycle. A
product life cycle of three months or less (98th
percentile) was coded as full membership in an
environment with a high rate of change; a product
life cycle of ten years (25th percentile) was coded
as full nonmembership in such an environment;
and a life cycle of three years (77th percentile) was
set as the crossover point.7
Although an environment may be changing rap-
idly, these changes may nevertheless be fairly pre-
dictable, and uncertainty is thus a second key di-
mension of an organization’s environment (Miles &
Snow, 1978). I assessed environmental uncertainty
using two items that asked how predictable
changes had been in the business environment over
the past three years. The two items assessed the
6
I use the term “rate of change” to distinguish this
construct from the related but more extensive and mul-
tidimensional construct “environmental velocity” (Mc-
Carthy, Lawrence, Wixted, & Gordon, 2010).
7
A lack of information on each firm’s industry rather
than overall sector prevented my measurement of envi-
ronmental uncertainty and rate of change using archival
measures at the industry level. However, recent studies
indicate that the survey-based measures of environmen-
tal characteristics used here are highly correlated with
alternative archival measures (Baum & Wally, 2003; Men-
delson & Pillai, 1999), suggesting that they are a satisfac-
tory choice for the current purposes.
turing products and technology related to product
improvement (1 ⫽ “easily predictable” to 5 ⫽
“completely unpredictable”). Both items were com-
bined into a scale that showed very good reliability
(␣ ⫽ .81). The fuzzy set measure of high environ-
mental uncertainty was based on this scale and
coded as fully out of the set for values of 1 (“easily
predictable”). Because the maximum observed
value was 4, I coded this value as fully in the set of
high environmental uncertainty and used the ob-
served scale midpoint of 2.5 as the crossover point.
Finally, as suggested by theory, the measures of rate
of change and uncertainty tap into different con-
structs, as shown by their low and nonsignificant
correlation of .05 (p ⫽ 0.56).8
Information on the survey items used to con-
struct the independent measures was missing for
an average of 14 percent of cases, and listwise de-
letion would have significantly reduced the overall
sample size and likely resulted in biased results
(Little & Rubin, 1987). I therefore imputed the miss-
ing values using maximum-likelihood estimation
based on information from all measures (Schafer,
1997). However, I did not impute missing values
for my outcome measure, and deleting cases with
missing performance information resulted in a
valid sample size of 139 cases. All subsequent anal-
yses refer to this final sample.
Calibration
As described above, the process of transforming
variables into sets requires the specification of full
membership in a set of interest, full nonmember-
ship, and a crossover point of maximum ambiguity
regarding membership. Given these three qualita-
tive anchors, one can transform variable raw scores
into set measures using the direct method of cali-
bration described by Ragin (2008). The basic intu-
ition underlying this calibration is that it rescales
an interval variable using the crossover point as an
anchor from which deviation scores are calculated,
taking the values of full membership and full non-
8
Environmental uncertainty is arguably a multifac-
eted construct that can also relate to, for example, polit-
ical uncertainty and uncertainty about input supply and
factor prices, and my measure may thus not capture the
full spectrum of the entrepreneurial problem—or the
broadness versus narrowness of a firm’s product-market
domain (Miles & Snow, 2003). However, the current mea-
sure of technological uncertainty is arguably most closely
related to the strategies of cost leadership and differenti-
ation and consistent with prior work on the typology
(e.g., Desarbo et al., 2005).
2011
membership as the upper and lower bounds.9 The
rescaled measures range from 0 to 1, and the con-
verted scores are tied to the thresholds of full mem-
bership, full nonmembership, and the crossover
point. In the current version of the fs/QCA software
package (2.5), the transformation is automated in
the “compute” command and can be easily exe-
cuted once the three thresholds are defined.
RESULTS
Table 3 presents descriptive statistics and corre-
lations for all measures, including some measures
used in supplementary analyses shown in the Ap-
pendix. The table shows the expected positive cor-
relations between size, formalization, and adminis-
trative complexity. In contrast, centralization is
negatively correlated with these three measures,
which is consistent with the notion that smaller
organizations with few levels of hierarchy concen-
trate decision making at the executive level. As
would be expected, there is also a significant neg-
ative correlation between environmental uncer-
tainty and the cost leadership strategy.
High-Performance Configurations
Table 4 shows the results of my fuzzy set analysis
of high performance. I use the notation for solution
tables recently introduced by Ragin and Fiss
(2008), according to which black circles (“●”) in-
dicate the presence of a condition, and circles with
a cross-out (“R”) indicate its absence. Furthermore,
large circles indicate core conditions, and small
circles refer to peripheral conditions. Blank spaces
in a solution indicate a “don’t care” situation in
which the causal condition may be either present
or absent.10 Solutions are grouped by their core
conditions.
9
An intermediate step of the direct method of calibra-
tion involves the transformation of these deviation scores
into the metric of log odds, which is advantageous since
this metric is centered around 0 and has no upper or
lower bound. For a detailed description of the calibration
procedure, see Ragin (2008: 86–94). Because the laws
governing the intersection of fuzzy sets make cases with
scores of exactly 0.5 difficult to analyze, Ragin (2008)
recommended avoiding the use of a precise 0.5 member-
ship score for causal conditions. To achieve this, I added
a constant of 0.001 to the causal conditions below full
membership scores of 1. Adding this constant to all con-
ditions does not affect the results of the regression anal-
yses reported in the Appendix but does assure that no
cases are dropped from the fuzzy set analyses.
10
The solution tables only list configurations that con-
sistently led to the outcome of interest; the tables do not
Fiss 407
The solution table shows that the fuzzy set
analysis results in four solutions exhibiting ac-
ceptable consistency (ⱖ 0.80) and furthermore
indicates the presence of both core and periph-
eral conditions as well as neutral permutations of
two configurations. The presence of several over-
all solutions thus points to a situation of first-
order, or across-type, equifinality of solutions,
and the neutral permutations within solutions 1
(1a and 1b) and 3 (3a and 3b) furthermore point to
the existence of second-order, or within-type,
equifinality.
Regarding core conditions, solutions 1a and 1b
indicate that a cost leadership strategy combining
formalization and centralization as well as the ab-
sence of uncertainty as peripheral conditions is
sufficient for achieving high performance, a profile
that fits the Defender type. These solutions further-
more suggest that, with a cost leadership strategy,
there are trade-offs between a high degree of com-
plexity and a high rate of environmental change.
Specifically, solution 1b of Table 4 indicates that
greater complexity allows for a firm’s high perfor-
mance regardless of whether its environment
changes at high speeds or not, as indicated by the
blank space for environmental change that signals a
“don’t care” situation for that causal condition. In
contrast, solution 1a shows the opposite pattern: in
the absence of a high rate of change, complexity
may be either high or low. Comparing solutions
1a and 1b thus indicates that high complexity
and the absence of a high rate of change can be
treated as substitutes. Furthermore, both solu-
tions 1a and 1b show that not being large and
using a differentiation strategy at least to some
extent are also parts of this causal configuration.
Interestingly, this finding lends some support to
Walker and Ruekert (1987), who argued for consid-
ering the notion of a differentiated Defender firm
that aims to protect a niche while also pursuing
some differentiation. Finally, note how the current
findings highlight the ability of QCA to explain the
relationships internal to configurations, and partic-
ularly substitution and complementarity effects
that usually remain in a black box in the more
standard statistical approaches.
Solution 2 indicates the existence of a success-
ful hybrid configuration that combines differen-
tiation and cost leadership as core conditions and
exhibits low formalization. This solution resem-
include configurations that do not lead to high perfor-
mance, that did not pass the frequency threshold, or that
showed no consistent pattern and thus did not pass the
consistency threshold.
408 Academy of Management Journal April

TABLE 3
Descriptive Statistics and Correlationa

Variable Mean s.d. 1 2 3 4 5 6 7 8 9 10 11 12 13

1. Size 0.31 0.29

2. Formalization 0.62 0.24 .35
3. Centralization 0.52 0.24 ⫺.20 ⫺.11
4. Administrative complexity 0.51 0.25 .53 .42 ⫺.20
5. Differentiation strategy 0.66 0.30 ⫺.02 .18 ⫺.02 .14
6. Cost leadership strategy 0.40 0.25 .09 .12 .05 .14 ⫺.02
7. Environmental rate of change 0.35 0.30 .04 ⫺.03 ⫺.01 .13 .01 .17
8. Environmental uncertainty 0.41 0.29 .03 ⫺.19 ⫺.06 ⫺.20 ⫺.17 ⫺.27 .05
9. Ideal type fit (minimum) ⫺1.67 0.35 ⫺.10 ⫺.34 .31 ⫺.26 ⫺.07 ⫺.07 ⫺.04 .10
10. Prospector deviation 2.38 0.59 .66 .55 .17 .61 ⫺.32 .49 .10 ⫺.16 ⫺.13
11. Analyzer deviation 1.79 0.44 ⫺.45 ⫺.22 .15 ⫺.67 .08 ⫺.22 ⫺.05 .12 ⫺.24 ⫺.54
12. Defender deviation 2.62 0.59 ⫺.66 ⫺.55 ⫺.17 ⫺.61 .32 ⫺.49 ⫺.10 .16 .13 ⫺1.00 .54
13. High performance 0.69 0.43 ⫺.06 ⫺.10 .11 ⫺.01 .09 .03 .11 ⫺.07 .12 ⫺.05 ⫺.05 .05
14. Very high performance 0.64 0.43 ⫺.08 ⫺.08 .13 ⫺.03 .12 .04 .13 ⫺.09 .12 ⫺.06 ⫺.03 .06 .97

a
Correlations of 0.17 or higher are significant at ⱕ.05.

bles the Analyzer type of Miles and Snow but, in
contrast to their theory, it appears that this type
does not operate well in either quickly changing
or uncertain environments.
Solutions 3a and 3b indicate a third important
path to high performance, combining a differentia-
tion strategy with informal organization, which is
consistent with a Prospector profile. Indeed, in
terms of structure and strategy, solution 3a is per-
fectly consistent with the Prospector ideal profile
defined by Miles and Snow when peripheral con-
ditions are taken into account. Solution 3b differs
slightly from 3a in that this solution combines com-
plexity with operating in a rapidly changing envi-

TABLE 4
Configurations for Achieving High Performancea

Solution

Configuration 1a 1b 2 3a 3b 4

Structure
Large size R R R R R 䢇
Formalization ● ● R R R ●
Centralization ● ● ● R R R
Complexity ● R ● R ●

Strategy
Differentiation ● ● 䢇 䢇 䢇 ●
Low cost 䢇 䢇 䢇 R R

Environment
Rate of change R R ● R R
Uncertainty R R R R R R

Consistency 0.82 0.82 0.86 0.83 0.83 0.82

Raw coverage 0.22 0.22 0.17 0.14 0.19 0.19
Unique coverage 0.01 0.01 0.02 0.01 0.02 0.04

Overall solution consistency 0.80

Overall solution coverage 0.36

a
Black circles indicate the presence of a condition, and circles with “⫻” indicate its absence. Large circles indicate core conditions;
small ones, peripheral conditions. Blank spaces indicate “don’t care.”
2011 Fiss 409

ronment as peripheral conditions. Note that for all TABLE 5

solutions except solution 4, the environment is Configurations for Achieving Very High Performancea
not a core condition, although higher levels of
Solution
uncertainty appear to hinder high returns for all
configurations at this performance level. Still, in Configuration 1a 1b 2
line with the Miles and Snow assumption, an
informally organized Prospector configuration Structure
(solution 3a) is better positioned to operate in a Large size ● R R
quickly changing environment than any other Formalization ● ● R
Centralization R ● ●
configuration.
Finally, solution 4 indicates that, in fairly pre-
Complexity 䢇 䢇 R

dictable environments, size also allows firms to Strategy

achieve high returns, which indicates the existence Differentiation ● ● 䢇
of economies of scale. However, these economies Low cost 䢇 䢇 ●
appear to be highly dependent upon a stable indus-
try environment, as indicated by the core condition Environment
requiring that environmental uncertainty be “not Rate of change R R R
high.” Uncertainty R R 䢇
The table also lists coverage scores that indicate the
percentage of cases that take a given path to the out- Consistency 0.83 0.83 0.84
come, allowing me to evaluate the importance of dif- Raw coverage 0.17 0.22 0.17
ferent causal paths. In terms of overall coverage, the Unique coverage 0.03 0.04 0.03
combined models account for about 36 percent of
membership in the outcome. Although this value is
Overall solution consistency 0.81
substantive, it also indicates considerable elements of Overall solution coverage 0.27
randomness or idiosyncrasy within configurations
a
that lead to high performance. Finally, the models in Black circles indicate the presence of a condition, and cir-
Table 4 indicate the existence of two possible neces- cles with “⫻” indicate its absence. Large circles indicate core
conditions; small ones, peripheral conditions. Blank spaces in-
sary conditions that are shared across all solutions,
dicate “don’t care.”
namely a lack of uncertainty and a differentiation
strategy. However, since the solutions do not cover all
possible paths to achieving high performance, and thus also indicating the presence of second-order
since there are in fact other configurations that do not equifinality. Note also that the overall number of
pass the consistency and frequency thresholds im- solutions has dropped from five to three and that,
posed here but do lead to high performance, the re- for all solutions in Table 5, the minimum number
sults indicate the existence of several sufficient solu- of core conditions has increased from one to
tions but likely no necessary condition for achieving three, indicating the existence of fewer choices
high performance in this sector. These findings dem- with greater constraints when aiming for very
onstrate the ability of a set-theoretic approach to ex- high performance, which provides evidence of
amine the necessity and sufficiency of configurations asymmetric causality.
and their elements, conditions that are not easily ex- The table also shows the existence of a highly
amined using standard, non-Boolean approaches. successful Prospector configuration in solution 2,
which largely resembles the Prospector configura-
tion of the previous table but now includes uncer-
Configurations for Very High Performance
tainty as a core condition. This finding is consistent
Table 5 shows the results for a fuzzy set analysis with the Prospector prototype of a small firm with
of very high performance. The results indicate the informal relations and centralized decision making
existence of two distinct configurational groupings, pursuing a differentiation strategy only in a highly
which again suggests the presence of first-order uncertain but not-too-quickly changing environ-
equifinality. Solutions 1a and 1b again rely on a ment. It also indicates that very high performance
cost leadership strategy in combination with a is possible even in normally unfavorable environ-
high degree of complexity and avoidance of rap- ments, given the right configuration.
idly changing environments. The solutions also Note further that the results of Table 5 indicate
show clear trade-offs, with size and centraliza- that no hybrid Analyzer configuration achieves
tion substituting for each other and allowing for very high performance. This is an important find-
neutral permutations around the core conditions, ing, as it indicates that trade-offs may become in-
410 Academy of Management Journal
creasingly pronounced as one moves up the perfor-
mance scale. The results thus indicate that it may
be possible to achieve high performance using a
hybrid type, but as one approaches very high per-
formance, trade-offs between differentiation and
cost leadership as well as their associated charac-
teristics of organizational structure appear to make
hybrid types such as the Analyzer infeasible: the
very high performers appear to rely on pure types.
In terms of coverage, the solutions account for
about 27 percent of membership in the group achiev-
ing very high performance, which is slightly less than
for the analysis of high performance above. Although
there are thus three sufficient configurations, the
analyses again indicate the same potentially neces-
sary conditions with the addition of a low-cost strat-
egy, but the less-than-perfect coverage and examina-
tion of conditions not passing the minimum
threshold test again suggest that there are other paths
to the outcome in question that do not rely on these
conditions.
Configurations for Not-High or Low Performance
These differences between configurations lead-
ing to high versus very high performance already
demonstrate the need to shift toward an asymmet-
ric understanding of causality. However, to further
explore this issue, I also conducted fuzzy set anal-
yses modeling the absence of high performance as
well as low performance. Note again that with stan-
dard regression analyses, this kind of analysis (i.e.,
predicting the absence of high performance) is al-
ways part of the process because of the symmetry of
relationships in regression models. However, in the
theoretical perspective suggested here, causal con-
ditions leading to the presence of an outcome may
frequently be different from those conditions lead-
ing to the absence of the outcome.
In line with an asymmetric understanding of cau-
sality in configurations, a fuzzy set analysis of the
absence of high performance indicated no consis-
tently identifiable solution, and consistency scores
for all solutions remained considerably below the
acceptable level of 0.75. These findings indicate the
absence of a clear set-theoretic relationship when
either the absence of high performance or the pres-
ence of low performance is used as the outcome. In
other words, there are many ways to be nonper-
forming here, but no consistent pattern.
In combination with the results above, modeling
the absence of the outcome complements the current
findings to suggest a clear picture of asymmetric cau-
sality. Specifically, the current analyses describe the
results for four different outcomes: very high perfor-
mance (ROA of 25%), high performance (ROA of
April
16.3%), not-high performance (ROA of 7.8%), and
low performance (ROA of 0.0%). The results indicate
that few configurations consistently lead to high per-
formance, and even fewer consistently lead to very
high performance, but no configuration of strategy,
structure, and environment consistently leads to av-
erage or below-average performance.
Sensitivity Analyses
I conducted several robustness checks and sensi-
tivity analyses. First, I compared the results of the
fsQCA analyses conducted here with results of
more traditional methods for the analysis of typol-
ogies, such as cluster analysis (e.g., Ketchen et al.,
1993) and deviation score approaches (e.g., Doty et
al., 1993). The Appendix provides details on esti-
mation and results. These provide broad support
for the existence of the Miles and Snow types and
their relationship with performance, yet they also
show the methodological differences of these ap-
proaches that allow for a limited insight into the
causal processes inside typologies. Specifically, al-
though the results of the cluster and deviation score
analyses support the overall typology, in contrast to
my fuzzy set QCA, they offer only limited insight
into the internal causal structure of the different
types, the neutral permutations inherent to the
types, and the asymmetric causal relationships
present across the performance spectrum.
Furthermore, I conducted sensitivity analyses to
examine whether my findings are robust to the use
of alternative specifications of my causal condi-
tions, using a different coding for complexity, size,
and the rate of change, where alternative crossover
points would appear to be most plausible. Specifi-
cally, I varied the crossover point between ⫹/–25
percent for all three measures. Minor changes are
observed regarding the kind of neutral permuta-
tions that occur as well as the specific number of
solutions and subsolutions, but the interpretation
of the results remains substantively unchanged.
DISCUSSION
Although typologies have figured prominently in
organization and strategy research and retain their
attractiveness, as evidenced by a number of recent
studies, their promise as well as that of configura-
tional theory still remains unfulfilled (cf. Short et
al., 2008). In this study, I have argued that a key
challenge of typological theory relates to understand-
ing the cause-effect relationships inherent to config-
urations. To overcome this challenge and allow the
building of better causal theories, I have proposed a
theoretical perspective that shifts the focus toward a
2011
definition of configurational core and periphery
based on causal relations with an outcome. Such a
shift allows a finer-grained understanding of typolog-
ical theories by additionally introducing the concepts
of neutral permutations and causal asymmetry.
In addition to introducing a fresh theoretical per-
spective and a vocabulary for understanding cause-
effect relationships in typologies, I introduced
fuzzy set analysis as a corresponding method for
more clearly understanding just what elements of a
configuration are relevant for an outcome and how
these elements combine to achieve their effects.
Specifically, the solutions found for the sample of
high-technology firms examined here demon-
strated the existence of several equifinal configura-
tions that included core and peripheral elements as
well as neutral permutations of these elements. In
this regard, the set-theoretic methods used here
hold considerable promise for overcoming the cur-
rent challenges and allow for a detailed analysis of
the necessary and sufficient conditions of high-
performance configurations. In combining a theoret-
ical approach and a novel methodology, the current
study thus represents a step toward building a better
understanding of the crucial role of cause-effect rela-
tionships in organizations, a theme that is central to
both the strategy and organization literatures.
As I have noted, the set-theoretic methods used
here allow for the analysis of causal asymmetry—
that is, they take into account the fact that the
configurations leading to very high performance
are frequently different from those leading to
merely high or average performance. So far, causal
asymmetry has for the most part been neglected in
both typological theory and organizational research
more broadly. However, causal asymmetry is argu-
ably pervasive in both domains, and failing to take
this causal structure into account is likely to lead to
incomplete or incorrect recommendations. Specif-
ically, the analysis of causal asymmetry in the cur-
rent study showed that firms with hybrid Analyzer
configurations combining elements of both Pros-
pector and Defender were indeed able to achieve
high performance, but such hybrids were not able
to achieve very high performance. Instead, config-
urations exhibiting very high performance resem-
bled the pure types rather than hybrids, apparently
indicating that achieving very high performance
means embracing trade-offs between elements.
This finding carries direct implications for the
growing literature on “organizational ambidexter-
ity” (e.g., Tushman & O’Reilly, 1996). Specifically,
shifting toward an asymmetric understanding of
how ambidexterity relates to performance may re-
solve some of the mixed findings on that relation-
ship (cf. Raisch & Birkinshaw, 2008).
Fiss 411
The notion of causal asymmetry also carries im-
plications for strategy and organization research
more broadly. Specifically, most current research
appears to imply a linear (or curvilinear) relation-
ship between its theoretical constructs of interest,
leading to a potential mismatch between an essen-
tially symmetrical theoretical relationship and an
actual underlying asymmetric causal relationship.
If so, this mismatch may be to blame for the incon-
sistent empirical findings that have plagued several
literatures, such as that on the relationship between
strategic change and firm performance (e.g., Rajago-
palan & Spreitzer, 1997) and that on the relation-
ship between corporate governance practices and
performance (e.g., Daily, Dalton, & Cannella, 2003).
For instance, consider the possibility that good gov-
ernance is a necessary though not sufficient condi-
tion for high performance. If this is the case, then
firms will on average need good governance ar-
rangements to maintain high performance, but by
themselves such governance arrangements will not
guarantee high performance. Although perfectly
consistent with a set-theoretic, asymmetric relation-
ship, such a pattern of data would in fact lead to weak
or no correlations between good governance and per-
formance. If this is correct, then applying the theo-
retic approach and methods used here may hold con-
siderable promise for resolving inconsistent findings.
Furthermore, the perspective and methods em-
ployed here allow for a close analysis of the equi-
final configurations leading to both high and very
high performance. By incorporating the notion of
neutral permutations into researchers’ understand-
ing of causal configurations, I have argued that
different constellations of peripheral elements that
are equifinal regarding the outcome in question
may surround the causal core of a configuration. To
refine the concept of equifinality, I have introduced
the notions of first-order equifinality (i.e., equifi-
nality across types) and second-order equifinality
(i.e., equifinality of permutations within types).
The distinction between these forms is important
for at least two reasons. First, the different types
and permutations may be equifinal for one outcome
but not for other relevant outcomes. Different types
or permutations may, for instance, result in differ-
ent interactions with other organizational or envi-
ronmental characteristics. Second, different types
and permutations are likely to affect future organi-
zational states in affecting the trajectories of subse-
quent organizational development by establishing
“path dependencies,” thus making certain trajecto-
ries more likely while reducing the likelihood of
others. The current study thus carries important
implications for the organizational design literature
by providing not only insight into the workings of
412 Academy of Management Journal
design elements but also a way to conceptualize
how such elements will affect future organizational
change efforts (Grandori & Furnari, 2008; Rivkin &
Siggelkow, 2007). In addition, although I have not
focused here on change in configurations, future re-
search will hopefully expand toward understanding
dynamic mechanisms in configurations or the notion
that configurations themselves may be dynamic.
A further aim of this study has been to demon-
strate the added value that a fuzzy set analysis
using QCA can bring to the study of typologies,
both in terms of better explanation of how causes
combine to create an outcome and more direct
modeling of equifinality in organizational configu-
rations. In this regard, fsQCA is a particularly use-
ful tool for understanding both complementarities
and substitutes in configurations. Accordingly,
fuzzy set analysis appears to complement other
standard approaches to testing typological theo-
ries—such as cluster analysis and deviation score
approaches— by enhancing understanding of mid-
range theories of the causal processes involved in
configurations rather than grand theories of overall
types. It is to this fine-grained examination of
causal processes as well as causal asymmetry that
fuzzy set approaches can particularly contribute.
Naturally, the current study also has limitations.
As Miller (1986) noted, the concepts of strategy,
structure, and environment are quite broad and
involve multiple dimensions. Accordingly, any
study aimed at examining typologies crossing these
three domains can only select a representative set
of categories for characterizing each of the do-
mains. The current study is no exception in that it
focused on some measures to the exclusion of oth-
ers that could be used to characterize variables
such as environment. Nevertheless, the measures
selected for this study are arguably central to the
three domains examined here, and the study is
quite comprehensive in that it is one of only a
handful to simultaneously include measures of the
structure, strategy, and environment relating to a ty-
pology. As such, it goes beyond much previous work
in offering a holistic assessment of typological con-
figurations across a multidimensional property space.
The limited sample size of the current study did
not permit further statistical testing for the fuzzy set
analyses. Although fuzzy set QCA can generally em-
ploy significance tests to examine, for instance, the
consistency of a solution, the specific causal structure
of the current sample, which included a number of
viable solutions, resulted in too few cases for each
11
Marx (2010) provides some guidance regarding the
appropriate ratio of causal conditions to cases.
April
solution to permit statistical tests. This result neces-
sarily limits the ability to draw definite conclusions
from this data set and calls for further studies to verify
the current results. Similarly, the current study was
able to draw on cross-industry data, yet the findings
are restricted to the high-technology sector. Although
this sector includes a number of important and highly
relevant industries such as semiconductors, com-
puter equipment, and airplane manufacturing, future
research should naturally aim to expand its scope
beyond the current empirical setting. Likewise, it
would be preferable to have further information re-
garding the representativeness of the sample relative
to the overall population of U.K. high-tech firms and
to expand the analysis to the industry level rather
than the aggregate high-tech sector. However, al-
though the findings of the current study are thus
limited in their generalizability, the logic of conclu-
sions is not context-specific and thus offers ample
opportunity for more research.
QCA appears particularly appropriate for the study
of complex causal relationships and multiple interac-
tions, yet this ability also has limitations that may
make the method more appropriate for some contexts
than others. In particular, because QCA is based on
fully interactive models that take all possible config-
urations into account, its data matrices increase ex-
ponentially with the number of causal conditions
considered. Accordingly, the number of cases avail-
able restricts the number of causal conditions ana-
lyzed simultaneously, and a researcher needs to be
careful to assure there are sufficient degrees of free-
dom to avoid the results being overdetermined.11
Typologies are likely to continue playing a central
role in management and strategy research, not the
least because the configurations embedded in them
arguably present the essence of strategy and are likely
to be a far greater source of competitive advantage
than any single aspect of an organizational system
(Miller, 1986: 510). In the current study I have argued
that the theory of typologies might benefit both con-
ceptually and empirically from a reorientation to-
ward the concepts of causal core and periphery, neu-
tral permutations, and causal asymmetry. I hope that
I have made a case for more research to extend this
approach and further show its utility in developing
the theory of causal mechanisms in organizations.
REFERENCES
Abrahamson, E., & Fombrun, C. 1994. Macro-cultures:
Determinants and consequences. Academy of Man-
agement Review, 19: 728 –755.
Abrahamson, E., & Rosenkopf, L. 1997. Social network
effects on the extent of innovation diffusion: A
2011
computer simulation. Organization Science, 8:
289 –309.
Althusser, L. 1972. Politics and history: Montesquieu,
Rousseau, Hegel and Marx [B. Brewster, trans.]).
London: NLB.
Ashmos, D. P., & Huber, G. P. 1987. The systems para-
digm in organization theory: Correcting the record
and suggesting the future. Academy of Manage-
ment Review, 12: 607– 621.
Bailey, K. D. 1973. Constructing monothetic and poly-
thetic typologies by the heuristic method. Sociolog-
ical Quarterly, 14: 291–308.
Bakker, R. M., Cambré, B., Korlaar, L., & Raab, J. 2010.
Managing the project learning paradox: A set-theo-
retic approach toward project knowledge transfer.
International Journal of Project Management; In
press.
Barr, P. S., Stimpert, J. L., & Huff, A. S. 1992. Cognitive
change, strategic action, and organizational renewal.
Strategic Management Journal, 13: 15–36.
Baum, J. R., & Wally, S. 2003. Strategic decision speed
and firm performance. Strategic Management Jour-
nal, 24: 1107–1129.
Bendix, R. 1963. Concepts and generalizations in com-
parative sociological studies. American Sociologi-
cal Review, 28: 532–538.
Blau, P. M., & W. R. Scott. 1962. Formal organizations.
San Francisco: Chandler.
Bogner, W. C., & Barr, P. S. 2000. Making sense in hyper-
competitive environments: A cognitive explanation
for the persistence of high velocity competition. Or-
ganization Science, 11: 212–226.
Borgatti, S. P., & Everett, M. G. 1999. Models of core
periphery structures. Social Networks, 21: 375–395.
Bourgeois, L. J., III, & Eisenhardt, K. M. 1988. Strategic
decision processes in high velocity environments.
Management Science, 34: 816 – 835.
Burns, T., & Stalker, G. M. 1961. The management of
innovation. London: Tavistock.
Burton, R. M., & Obel, B. 2004. Strategic organizational
diagnosis and design: The dynamics of fit (3rd ed.).
New York: Springer.
Capecchi, V. 1966. Typologies in relation to mathemati-
cal models. Ikon, 19: 63–124.
Carley, K., & Palmquist, M. 1992. Extracting, representing
and analyzing mental models. Social Forces, 70:
601– 636.
Child, J. 1972.Organization structure and strategies of
control: A replication of the Aston study. Adminis-
trative Science Quarterly, 17: 163–177.
Cosh, A. D., Hughes, A., Gregory, M., & Jayanthi, S. 2002.
Cambridge Centre for Business Research Manu-
facturing strategy and competitiveness data set,
1994 –1999 (computer file). Colchester, U.K.: Data
Fiss 413
Archive (distributor). SN: 4434. Retrieved March
2002.
Crilly, D. In press. Predicting stakeholder orientation in
the multinational corporation: A mid-range theory.
Journal of International Business Studies.
Daily, C. M., Dalton, D. R., & Cannella, A. A. 2003. Cor-
porate governance: Decades of dialogue and data.
Academy of Management Review, 28: 371–382.
David, J. S., Hwang, Y., Pei, B. K. W., & Reneau, J. H.
2002. The performance effects of congruence be-
tween product competitive strategies and purchas-
ing management design. Management Science, 48:
866 – 885.
Delery, J. E., & Doty, D. H. 1996. Modes of theorizing in
strategic human resource management: Tests of uni-
versalistic, contingency, and configurational perfor-
mance predictions. Academy of Management Jour-
nal, 39: 802– 835.
DeSarbo, W. S., Di Benedetto, C. A., Song, M., & Sinha, I.
2005. Revisiting the Miles and Snow strategic frame-
work: Uncovering interrelationships between strate-
gic types, capabilities, environmental uncertainty,
and firm performance. Strategic Management Jour-
nal, 26: 47–74.
DeSarbo, W. S., Grewal, R., Wang, R. 2009. Dynamic
strategic groups: Deriving spatial evolutionary paths.
Strategic Management Journal, 30: 1420 –1439.
Dess, G. G., & Beard, D. W. 1984. Dimensions of organi-
zational task environments. Administrative Science
Quarterly, 29: 52–73.
Doty, D. H., & Glick, W. H. 1994. Typologies as a unique
form of theory building: Toward improved under-
standing and modeling. Academy of Management
Review, 19: 230 –251.
Doty, D. H., Glick, W. H., & Huber, G. P. 1993. Fit,
equifinality, and organizational effectiveness: A test
of two configurational theories. Academy of Man-
agement Journal, 36: 1196 –1250.
Dranove, D., Peteraf, M., & Shanley, M. 1998. Do strategic
groups exist? An economic framework for analysis.
Strategic Management Journal, 19: 1029 –1044.
Durand, R., & Vaara, E. 2009. Causation, counterfactuals,
and competitive advantage. Strategic Management
Journal, 30: 1245–1264.
Eden, C., Ackermann, F., & Cropper S. 1992. The analysis
of cause maps. Journal of Management Studies, 29:
309 –324.
Etzioni, A. 1961. A comparative analysis of complex
organizations. New York: Free Press.
Ferguson, T. D., Deephouse, D. L., & Ferguson, W. L.
2000. Do strategic groups differ in reputation? Stra-
tegic Management Journal, 21: 1195–1214.
Fine, C. H. 1998. Clockspeed: Winning industry control
in the age of temporary advantage. Reading, MA:
Perseus.
414 Academy of Management Journal
Fiss, P. C. 2007. A set-theoretic approach to organization-
al configurations. Academy of Management Re-
view, 32: 1180 –1198.
Fiss, P. C. 2009. Case studies and the configurational
analysis of organizational phenomena. In C. Ragin &
D. Byrne (Eds.), Handbook of case study methods:
424 – 440. Thousand Oaks, CA: Sage.
Ford, J. D. 1985. The effects of causal attributions on
decision makers’ responses to performance down-
turns. Academy of Management Review, 10: 770 –
786.
Galaskiewicz, J., & Wasserman, S. 1989. Mimetic pro-
cesses within an interorganizational field: An empir-
ical test. Administrative Science Quarterly, 34:
454 – 479.
Galbraith, C., & Schendel, D. 1983. An empirical anal-
ysis of strategy types. Strategic Management
Journal, 4: 153–173.
Garcı́a Muñiz, A. S., & Ramos Carvajal, C. 2006. Core/
periphery structure models: An alternative method-
ological proposal. Social Networks, 28: 442– 448.
Geletkanycz, M. A. 1997. The salience of “culture’s con-
sequence”: The effects of cultural values on top ex-
ecutive commitment to the status quo. Strategic
Management Journal, 18: 615– 634.
Gopnik, A., & Schulz, L. (Eds.). 2007. Causal learning:
Psychology, philosophy, and computation. Oxford,
U.K.: Oxford University Press.
Grandori, A., & Furnari, S. 2008. A chemistry of organi-
zations: Combinatory analysis and design. Organi-
zation Studies, 19: 459 – 485.
Greckhamer, T. 2011. Cross-cultural differences in com-
pensation level and inequality across occupations: A
set-theoretic analysis. Organization Studies, 32: 85–
115.
Greckhamer, T., Misangyi, V. F., Elms, H., & Lacey, R.
2008. Using qualitative comparative analysis in stra-
tegic management research: An examination of com-
binations of industry, corporate, and business-unit
effects. Organizational Research Methods, 11: 695–
726.
Gresov, C., & Drazin, R. 1997. Equifinality: Functional
equivalence in organization design. Academy of
Management Review, 22: 403– 428.
Gulati, R., & Gargiulo, M. 1999. Where do inter-organi-
zational networks come from? American Journal of
Sociology, 104: 1439 –1493.
Gustafson, L. T., & Reger, R. K. 1995. Using organization-
al identity to achieve stability and change in high
velocity environments. Academy of Management
Best Papers Proceedings, 464 – 468.
Hahn, U., & Chater, N. 1997. Concepts and similarity. In
K. Lamberts & D. Shanks (Eds.), Knowledge con-
cepts and categories: 43–92. Cambridge, MA: MIT
Press.
April
Hambrick, D. C. 1983. Some tests of the effectiveness and
functional attributes of Miles and Snow’s strategic
types. Academy of Management Journal, 26: 5–26.
Hambrick, D. C. 2003. On the staying power of defenders,
analyzers, and prospectors. Academy of Manage-
ment Executive, 17(4): 115–118.
Hambrick, D. C., Geletkanycz, M. A., & Fred-Levinson,
J. W. 1993. Top executive commitment to the status
quo: Some tests of its determinants. Strategic Man-
agement Journal, 14: 401– 418.
Hannan, M. T., Burton, M. D., & Baron, J. N. 1996. Inertia
and change in early years: Employment relations in
young, high-technology firms. Industrial and Cor-
porate Change, 5: 503–536.
Hannan, M. T., & Freeman, J. 1984. Structural inertia and
organizational change. American Sociological Re-
view, 49: 149 –164.
Herrmann, A. M., & Cronqvist, L. 2009. When dichoto-
mization becomes a problem for the analysis of mid-
dle-sized data sets. International Journal of Social
Research Methodology, 12: 33–50.
Hodgkinson, G. 1997. Cognitive inertia in a turbulent
market: The case of U.K. residential estate agents.
Journal of Management Studies, 34: 921–945.
Hofer, C. W., & Schendel, D. 1978. Strategy formulation:
Analytical concepts. St. Paul: West.
Huff, A. S. (Ed.) 1990. Mapping strategic thought.
Chichester, U.K.: Wiley.
Huff, A. S., & Jenkins, M. 2001. Mapping strategic
knowledge. London: Sage.
Hult, G. T. M., Ketchen, D. J., Cavusgil, S. T., & Calan-
tone, R. J. 2006. Knowledge as a strategic resource in
supply chains. Journal of Operations Management,
24: 458 – 475.
Hunn, E. 1982. The utilitarian factor in folk biological
classification. American Anthropologist, 84:
830 – 847.
Inkpen, A., & Choudhury, N. 1995. The seeking of
strategy where it is not: Towards a theory of strat-
egy absence. Strategic Management Journal, 16:
313–323.
Jackson, G. 2005. Employee representation in the board
compared: A fuzzy sets analysis of corporate gover-
nance, unionism, and political institutions. Industri-
elle Beziehungen, 12: 1–28.
Jurkovich, R. 1974. A core typology of organizational
environments. Administrative Science Quarterly,
19: 380 –394.
Kabanoff, B., & Brown, S. 2008. Knowledge structures of
prospectors, analyzers, and defenders: Content,
structure, stability, and performance. Strategic
Management Journal, 29: 149 –171.
Katz, D., & Kahn, R. L. 1978. The social psychology of
organizations (2nd ed.). New York: Wiley.
2011
Kelly, D., & Amburgey, T. L. 1991. Organizational in-
ertia and momentum: A dynamic model of strate-
gic change. Academy of Management Journal, 34:
591– 612.
Ketchen, D. J., & Shook, C. L. 1996. The application of
cluster analysis in strategic management research:
An analysis and critique. Strategic Management
Journal, 17: 441– 485.
Ketchen, D. J., Thomas, J. B., & Snow, C. C. 1993. Orga-
nizational configurations and performance: A com-
parison of theoretical approaches. Academy of Man-
agement Journal, 36: 1278 –1313.
Kim, P. H., Dirks, K. T., Cooper, C. D., & Ferrin, D. L.
2006. When more blame is better than less: The
implications of internal vs. external attributions for
the repair of trust after a competence- vs. integrity-
based trust violation. Organizational Behavior &
Human Decision Processes, 99: 49 – 65.
King, A. W., Zeithaml C. P. 2001. Competencies and
firm performance: Examining the causal ambiguity
paradox. Strategic Management Journal, 22:
75–99.
Kogut, B., MacDuffie, J. P, & Ragin, C. C. 2004. Prototypes
and strategy: Assigning causal credit using fuzzy
sets. European Management Review, 1: 114 –131.
Konrad, A. M., & Linnehan, F. 1995. Formalized HRM
structures: Coordinating equal employment oppor-
tunity or concealing organizational practices?
Academy of Management Journal, 38: 787– 820.
Lazarsfeld, P. F. 1937. Some remarks on typological pro-
cedure in social research. Zeitschrift für Sozial-
forschung, 6: 119 –139.
Lim, L. K. S., Acito, F., & Rusetski, A. 2006. Development
of archetypes of international marketing strategy.
Journal of International Business Studies, 37: 499 –
524.
Little, R. J. A., & Rubin, D. B. 1987. Statistical analysis
with missing data. New York: Wiley.
Long, J. S. 1997. Regression models for categorical and
limited dependent variables. Thousand Oaks, CA:
Sage.
Lyles, M. A., & Schwenk, C. R. 1992. Top management,
strategy and organizational knowledge structures.
Journal of Management Studies, 29: 155–174.
March, J. G. 1991. Exploration and exploitation in organ-
izational learning. Organization Science, 2(special
issue: Organizational learning): 71– 87.
Marlin, D., Ketchen, D. J., & Lamont, B. 2007. Equifi-
nality and the strategic group-performance rela-
tionship. Journal of Managerial Issues, 19: 208 –
232.
Marx, A. 2008. Limits to non-state market regulations: A
qualitative comparative analysis of the international
sport footwear industry and the Fair Labor Associa-
tion. Regulation & Governance, 2: 253–273.
Fiss 415
Marx, A. 2010. Crisp-set qualitative comparative analysis
(csQCA) and model specification: Benchmarks for
future csQCA applications. International Journal of
Multiple Research Approaches, 4: 138 –158.
McCarthy, I. P., Lawrence, T. B., Wixted, B., & Gordon, B.
2010. A multidimensional conceptualization of en-
vironmental velocity. Academy of Management Re-
view, 35: 604 – 626.
McKelvey, B. 1975. Guidelines for the empirical classifi-
cation of organizations. Administrative Science
Quarterly, 20: 509 –525.
McKelvey, B. 1982. Organizational systematics: Taxon-
omy, evolution, classification. Berkeley: University
of California Press.
McKinney, J. C. 1966. Constructive typology and social
theory. New York: Appleton-Century-Crofts.
McKinney, J. C. 1969. Typification, typologies, and soci-
ological theory. Social Forces, 48: 1–12.
McNamara, G., Deephouse, D. L., & Luce, R. A. 2003.
Competitive positioning within and across a strate-
gic group structure: The performance of core, sec-
ondary, and solitary firms. Strategic Management
Journal, 24: 161–181.
McPhee, R. D., & Scott Poole, M. S. 2001. Organiza-
tional structures and configurations. In F. Jablin &
L. Putnam (Eds.), The new handbook of organi-
zational communication: Advances in theory, re-
search, and methods: 503–543. Thousand Oaks,
CA: Sage.
Mendelson, H., & Pillai, R. R. 1999. Industry clockspeed:
Measurement and operational implications. Manu-
facturing & Service Operations Management, 1:
1–20.
Meyer, A. D., Tsui, A. S., & Hinings, C. R. 1993. Configura-
tional approaches to organizational analysis. Acad-
emy of Management Journal, 36: 1175–1195.
Miles, R. E., & Snow, C. C. 1978. Organizational strat-
egy, structure, and process. [A. D. Meyer, collabo-
rator; H. J. Coleman Jr., contributor]. New York:
McGraw Hill.
Miles, R. E., & Snow, C. C. 2003. Organizational strat-
egy, structure, and process [A. D. Meyer, collabora-
tor; H. J. Coleman Jr., contributor]. Stanford, CA:
Stanford Business Classics.
Mill, J. S. [1843]/2002. A system of logic: Ratiocinative
and inductive. Honolulu: University Press of the
Pacific.
Miller, D. 1986. Configurations of strategy and structure:
A synthesis. Strategic Management Journal, 7:
233–249.
Miller, D. 1996. Configurations revisited. Strategic Man-
agement Journal, 17: 505–512.
Mintzberg, H. 1979. The structuring of organizations: A
synthesis of the research. Englewood Cliffs, NJ:
Prentice-Hall.
416 Academy of Management Journal
Mintzberg, H. 1983. Structures in fives: Designing ef-
fective organizations. Englewood Cliffs, NJ: Pren-
tice-Hall.
Nadkarni, S., & Narayanan, V. K. 2007a. The evolution of
collective strategy frames in high- and low-velocity
industries. Organization Science, 18: 688 –710.
Nadkarni, S., & Narayanan; V. K. 2007b. Strategic sche-
mas, strategic flexibility, and firm performance: The
moderating role of industry clockspeed. Strategic
Management Journal, 28: 243–270.
Nunnally, J. C. 1978. Psychometric theory (2nd ed.).
New York: McGraw Hill.
Olson, E. M., & Slater, S. F., Hult, G. T. M. 2005. The
performance implications of fit among business strat-
egy, marketing organization structure, and strategic be-
havior. Journal of Marketing, 69(July): 49 – 65.
Pajunen, K. 2008. Institutions and inflows of foreign
direct investment: A fuzzy-set analysis. Journal of
International Business Studies, 39: 652– 669.
Parnell, J. A. 1997. Top management team fit in the U.S.
airline industry. British Journal of Management, 8:
175–181.
Payne, G. T. 2006. Examining configurations and firm
performance in a suboptimal equifinality context.
Organization Science, 17: 756 –770.
Peteraf, M., & Shanley, M. 1997. Getting to know you: A
theory of strategic group identity. Strategic Manage-
ment Journal, 18(Summer special issue): 165–186.
Pfeffer, J. 1976. Beyond management and the worker: The
institutional function of management. Academy of
Management Review, 1: 36 – 46.
Porac, J., Thomas, H., & Baden-Fuller, C. 1989. Compet-
itive groups as cognitive communities: The case of
Scottish knitwear manufacturers. Journal of Man-
agement Studies, 26: 397– 416.
Porac, J. F., & Rosa, J. A. 1996. In praise of managerial
narrowmindedness. Journal of Management In-
quiry, 1: 35– 42.
Porac, J. F., & Thomas, H. 1990. Taxonomic mental mod-
els in competitor definition. Academy of Manage-
ment Review, 15: 224 –240.
Porter, M. E. 1980. Competitive strategy. New York: Free
Press.
Porter, M. E. 1996. What is strategy? Harvard Business
Review, 74(6): 61–78.
Pugh, D. S., Hickson, D. J., & Hinings, C. R. 1968. Dimen-
sions of organization structure. Administrative Sci-
ence Quarterly, 13: 65–105.
Ragin, C. C. 1987. The comparative method: Moving
beyond qualitative and quantitative strategies.
Berkeley: University of California Press.
Ragin, C. C. 2000. Fuzzy set social science. Chicago:
University of Chicago Press.
Ragin, C. C. 2005. From fuzzy sets to crisp truth tables.
April
Working paper, University of Arizona, Tucson, at
http://www.compasss.org/Raginfztt_April05.pdf.
Ragin, C. C. 2006. Set relations in social research: Eval-
uating their consistency and courage. Political Anal-
ysis, 14: 291–310.
Ragin, C. C. 2008. Redesigning social inquiry: Fuzzy sets
and beyond. Chicago: University of Chicago Press.
Ragin, C. C., & Fiss, P.C. 2008. Net effects analysis versus
configurational analysis: An empirical demonstra-
tion. In C. C. Ragin (Ed.), Redesigning social in-
quiry: Fuzzy sets and beyond: 190 –212. Chicago:
University of Chicago Press.
Raisch, S., & Birkinshaw, J. 2008. Organizational ambi-
dexterity: Antecedents, outcomes, and moderators.
Journal of Management, 34: 375– 409.
Rajagopalan, N., & Spreitzer, G. M. 1997. Toward a theory
of strategic change: A multi-lens perspective and
integrative framework. Academy of Management
Review, 22: 48 –79.
Reger, R., & Huff, A. 1993. Strategic groups: A cognitive
perspective. Strategic Management Journal, 14: 103–
124.
Reger, R. K., & Palmer, T. B. 1996. Managerial categori-
zation of competitors: Using old maps to navigate
new environments. Organization Science, 7: 22–39.
Reynolds, P. D. 1971. A primer in theory construction.
Indianapolis: Bobbs-Merrill.
Rich, P. 1992. The organizational taxonomy: Definition and
design. Academy of Management Review, 17: 758 –
781.
Rihoux, B., & Ragin, C. C. (Eds.). 2009. Configurational
comparative methods: Qualitative comparative
analysis (QCA) and related techniques. Thousand
Oaks, CA: Sage.
Rivkin, J. W., & Siggelkow, N. 2007. Patterned interac-
tions in complex systems: Implications for explora-
tion. Management Science, 53: 1068 –1085.
Romanelli, E., & Tushman, M. L. 1994. Organizational
transformation as punctuated equilibrium: An em-
pirical test. Academy of Management Journal, 37:
1141–1166.
Romme, A. G. 2003. Making a difference: Organization as
design. Organization Science, 14: 558 –573.
Rosch, E. 1978. Principles of categorization. In E. Rosch &
B. B. Lloyd (Eds.), Cognition and categorization:
27– 48. Hillsdale, NJ: Erlbaum.
Rosch, R. 1981. Prototype classification and logical classi-
fication: The two systems. In E. Scholnick (Ed.), New
trends in cognitive representation: Challenges to Pi-
aget’s theory: 73– 86. Hillsdale, NJ: Erlbaum.
Rose, A. 1950. A deductive ideal-type method. American
Journal of Sociology, 56: 35– 42.
Schafer, J. L. 1997. Analysis of incomplete multivariate
data (book no. 72, Chapman and Hall series “Mono-
2011
graphs on Statistics and Applied Probability”). Lon-
don: Chapman and Hall.
Schneider, C. Q., & Wagemann, C. 2007. Qualitative
comparative analysis (QCA) und fuzzy sets: Ein
Lehrbuch für Anwender und jene, die es werden
wollen. Mit einem Vorwort von Charles Ragin.
[Qualitative comparative analysis (QCA) and fuzzy
sets: A guide for current and future users]. Opladen,
Germany: Barbara Budrich.
Schneider, M. R., Schulze-Bentrop, C., & Paunescu, M.
2009. Mapping the institutional capital of high-tech
firms: A fuzzy-set analysis of capitalist variety and
export performance. Journal of International Busi-
ness Studies, 41: 246 –266.
Schwenk, C. R. 1984. Cognitive simplification processes
in strategic decision-making. Strategic Manage-
ment Journal, 5: 111–128.
Scott, W. R. 1981. Organizations: Rational, natural, and
open systems. Englewood Cliffs, NJ: Prentice Hall.
Segev, E. 1989. A systematic comparative analysis and
synthesis of two business-level strategic typologies.
Strategic Management Journal, 10: 487–505.
Short, J. C., Payne, G. T., & Ketchen, D. J. 2008. Research
on organizational configurations: Past accomplish-
ments and future challenges. Journal of Manage-
ment, 34: 1053–1079.
Shortell, S. M., & Zajac, E. J. 1990. Perceptual and archival
measures of Miles and Snow’s strategic types: A com-
prehensive assessment of reliability and validity.
Academy of Management Journal, 33: 817– 832.
Siggelkow, N. 2002. Evolution towards fit. Administra-
tive Science Quarterly, 47: 25–159.
Siggelkow, N. 2003. Why focus? A study of intra-industry
focus effects. Journal of Industrial Economics, 51:
121–150.
Singh, J. V. 1986. Technology, size, and organizational
structure: A reexamination of the Okayama study data.
Academy of Management Journal, 29: 800 – 812.
Singh, J. V., House, R. J., & Tucker, D. J. 1986. Organiza-
tional change and organizational mortality. Admin-
istrative Science Quarterly, 31: 587– 611.
Slater, S. F., & Olson, E. M. 2000. Strategy type and
performance: The influence of sales force manage-
ment. Strategic Management Journal, 21: 813– 829.
Sloman, S. A. 2005. Causal models: How people think
about the world and its alternatives. New York:
Oxford University Press.
Snow, C. C., & Hrebiniak, L. G. 1980. Strategy, distinctive
competence, and organizational performance. Ad-
ministrative Science Quarterly, 25: 317–336.
Stadler, B. M. R., Stadler, P. F., Wagner, G. P., & Fontana,
W. 2001. The topology of the possible: Formal spaces
underlying patterns of evolutionary change. Journal
of Theoretical Biology, 213: 241–274.
Stuart, T. E., & Podolny, J. M. 1996. Local search and the
Fiss 417
evolution of technological capabilities. Strategic Man-
agement Journal, 17(summer special issue): 21–38.
Thompson, J. D. 1967. Organizations in action. New
York: McGraw-Hill.
Tushman, M. L., & O’Reilly, C. A., III. 1996. Ambidex-
trous organizations: Managing evolutionary and rev-
olutionary change. California Management Re-
view, 38(4): 8 –30.
Van Aken, J. E. 2005. Management research as a design
science: Articulating the research products of mode
2 knowledge production in management. British
Journal of Management, 16: 19 –36.
Van de Ven, A., & Drazin, R. 1985. The concept of fit in
contingency theory. In L. L. Cummings & B. M. Staw
(Eds.), Research in organizational behavior, vol. 7:
333–365. Greenwich, CT: JAI Press.
Waldman, M. R., & Hagmayer, Y., & Blaisdell, A. P. 2006.
Beyond the information given: Causal models in
learning and reasoning. Current Directions in Psy-
chological Science, 15: 307–311.
Walker, O. C., & Ruekert, R. W. 1987. Marketing’s role in
the implementation of business strategies: A critical
review and conceptual framework. Journal of Mar-
keting 51(July): 15–33.
Weber, M. 1949. The methodology of the social sciences
[E. A. Shils & H. A. Finch, trans.]. New York: Free
Press.
Wiersema, M. F., & Liebeskind, J. P. 1995. The effects of
leveraged buyouts on corporate growth and diversi-
fication in large firms. Strategic Management Jour-
nal, 16: 447– 460.
Winch, R. F. 1947. Heuristic and empirical typologies.
American Sociological Review, 12: 68 –75.
Wong, G. Y. Y., & Birnbaum-More, P. H. 1994. Culture,
context and structure: A test of Hong Kong banks.
Organization Studies, 15: 99 –123.
Zahra, S. A., & Pearce, J. A., II. 1990. Research evidence
on the Miles-Snow typology. Journal of Manage-
ment, 16: 751–768.
Zajac, E. J., & Shortell, S. M. 1989. Changing generic strat-
egies: Likelihood, direction, and performance implica-
tions. Strategic Management Journal, 10: 413– 430.
APPENDIX
Supplemental Analyses
Cluster Analysis
To derive an empirical taxonomy, I used a two-step clus-
ter analysis, which has been the dominant tool of analysis
for configurations and strategic groups (DeSarbo, Grewal, &
Wang, 2009; Ketchen and Shook, 1996). I took the four
structural and two strategy variables and then examined
how the derived configurations performed given differing
environments. In a first step, hierarchical cluster analyses
using Ward’s minimum variance method suggested a three-
418 Academy of Management Journal
cluster solution based on cutoff values and inspection of
dendrograms (Ferguson, Deephouse, Ferguson, 2000; Mar-
lin et al., 2007). After determining this three-cluster solu-
tion, I used K-means cluster analysis in a second step, with
the centroid values of the hierarchical analysis as “seeds”
(e.g., Lim, Acito, & Rusetski, 2006; Payne, 2006). To assure
comparability across variables, all measures were standard-
ized prior to the analysis. Results of the cluster analysis
with final cluster centers are presented in Table A1 and are
essentially stable across different clustering algorithms.
As the table shows, the solution includes three clusters
that map somewhat imperfectly onto the Miles and Snow
typology. Group 1 (n ⫽ 93) corresponds roughly to the
Prospector type, scoring low on size formalization, com-
plexity, and cost leadership, and high on a differentiation
strategy. At the other end of the continuum, group 3 (n ⫽
59) approximates the Defender type, scoring high on size,
formalization, and complexity. However, its score of 0.42
on cost leadership is only slightly higher than that of group
1, yet its differentiation score is only slightly lower. Fur-
thermore, again diverging from the ideal typology, central-
ization scores are high for Prospectors and low for Defend-
ers. Finally, group 2 (n ⫽ 41) does appear to fit the Analyzer
profile and mostly occupies a middle position except for
differentiation, in which this group has the lowest score of
the three groups.
The empirical results thus bear some resemblance to the
Miles and Snow ideal types of Prospector, Analyzer, and
Defender, although the fit is less than perfect. To examine
the relationship between these ideal types and perfor-
mance, I regressed indicator variables for cluster member-
ship on the performance measure, including interaction
terms for environmental rate of change and uncertainty. I
used two-limit tobit regression, which is the appropriate
model when the dependent variable is truncated (Long,
1997), as is the case with the fuzzy set measure that cali-
brates the measure by introducing cutoffs for full member-
ship and nonmembership in the set of high-performing
TABLE A1
Cluster Analysis Results
Final Cluster Centers
Configuration 1 2 3 sponded to full nonmembership, with medium scores
Structure
Size 0.13 0.22 0.64
Formalization 0.52 0.49 0.76
Centralization 0.62 0.52 0.39
Complexity 0.40 0.42 0.71
Strategy
Differentiation 0.82 0.19 0.75
Low Cost 0.37 0.42 0.42
n 93 41 59
April
firms.a In the models, the Analyzer type is used as the
omitted category. Table A2 presents results.
Model 1 shows the effect of Prospector and Defender
types on performance, and models 2 to 5 add interac-
tion terms for environmental rate of change and uncertainty
to examine whether the types perform better or worse in
these environments as predicted by the typology. As model
1 shows, only the Prospector type exhibits significantly
higher performance than the Analyzer type. Furthermore,
models 2 through 5 offer no evidence that the performance
of either the Prospector or Defender type depends on the
environment in which it operates. Alternative models using
the Defender type as the omitted category also showed no
performance differences or dependence of performance on
the environment for the Analyzer type. Because models
using the alternative measure of very high performance
resulted in essentially identical results, these models are
not reported here but are available from the author upon
request. In addition, note that model fit is less than desir-
able, with pseudo-R2 values between .04 and .05. In sum,
the models offer only very limited support for performance
differences between these types, and they offer no support
for performance being contingent upon the nature of the
environment, as specified by the theory.
Deviation Score Analyses
I follow prior studies in using deviation scores to test
the relationship between the fit with a theoretical typol-
ogy and performance (Delery & Doty, 1996; Doty et al.,
1993; Doty & Glick, 1994). Profile definition is based on
the Miles and Snow ideal types of Prospector, Analyzer,
and Defender. For each profile, fit is calculated as the
deviation of an organization from an ideal type and
across all attributes. From the fit scores, ideal profile fit is
calculated as the minimum deviation across the three
profiles, according to the following formula:
冉 冊
l
Fit IT ⫽ ⫺ minD io .
i ⫽ 1
Here, Dio is the distance between ideal type i and
organization o, and the formula takes the minimum of
this distance across all ideal types (Doty et al., 1993).
However, although Doty et al. used uncalibrated mea-
sures to create their ideal profiles, I used the fuzzy set
measures to assure the comparability of all three types
of analysis. Accordingly, high ideal profile scores cor-
responded to full membership, and low scores corre-
tied to the crossover threshold. Because the Prospector
and Defender ideal types specified by the Miles and
Snow typology are the opposite of each other, their
deviation scores are inversely correlated with each
a
I also estimated models that used OLS regression in
combination with the uncalibrated performance mea-
sure. The results were substantively identical, with the
exception that in these models the coefficient for the pros-
pector type is significant at the .01 instead of the .05 level.
2011 Fiss 419

TABLE A2
Tobit Regression Models of Cluster Configurations on Performancea

Independent Variable Model 1 Model 2 Model 3 Model 4 Model 5

Prospector 1.14* (0.53) 1.16† (0.70) 0.26 (0.77) 1.12* (0.53) 1.15* (0.53)
Defender 0.13 (0.52) 0.13 (0.52) 0.08 (0.52) 0.80 (0.72) 0.93 (0.81)
Environmental rate of change 1.15 (0.71) 1.18 (0.92) 1.03 (0.70) 1.82* (0.90) 1.16 (0.71)
Environmental uncertainty ⫺0.85 (0.70) ⫺0.85 (0.70) ⫺1.65 (0.92) ⫺0.83 (0.69) ⫺0.21 (0.83)
Prospector ⫻ rate of change ⫺0.07 (1.38)
Prospector ⫻ uncertainty 1.98 (1.41)
Defender ⫻ rate of change ⫺1.94 (1.45)
Defender ⫻ uncertainty ⫺1.91 (1.47)

Constant 0.79 (0.54) 0.78 (0.58) 1.20 (0.61) 0.57 (0.56) 0.49 (0.58)
Likelihood-ratio ␹2 11.66* 11.66* 13.72* 13.53* 13.43*
Pseudo-R2 0.04 0.04 0.05 0.05 0.04

a
Standard errors are in parentheses; n ⫽ 139.
†
p ⱕ .10
* p ⱕ .05

other. I again used two-limit Tobit regression to model
the relationship between ideal profile deviation and
performance. In the models, the Analyzer type is used
as the omitted category. Table A3 presents the results.
Model 1 shows the results of an organization’s overall
ideal profile fit across all profiles, thus testing whether a
firm performs better if its structural features resemble
any of the ideal types identified by the theory (Doty et al.,
1993). The fit coefficient is positive and marginally sig-
nificant, which indicates that such a fit is likely associ-
ated with higher performance. To examine whether fit
with any particular ideal type drive this finding, models
2 to 4 show the coefficient for deviation from the individual
Prospector, Analyzer, and Defender profiles. However,
these models indicate that it is not simply one profile that is
superior. Note also that the findings using this theoretically
derived typology differ from the empirically derived solu-
tion based on cluster analysis, in which only the Prospector
type showed increased performance.
Models 5 through 8 show the interaction between de-
viation from the three ideal types and environmental rate
of change and uncertainty. As suggested by a model of
environmental contingency, model 6 indicates that devi-
ation from the Prospector profile decreases performance
in high-uncertainty environments. The models did not
indicate any support for the dependence of ideal type fit
on the environmental rate of change. Alternative models
using the very high performance measure as the depen-
dent variable resulted in essentially identical models and
are therefore omitted here. Note also that model fit as
measured by the pseudo-R2 is again rather poor, with
values between .02 and .04 for the overall model. Com-

TABLE A3
Tobit Regression Models of Profile Fit and Deviation on Performancea

Independent Variable Model 1 Model 2 Model 3 Model 4 Model 5 Model 6 Model 7 Model 8

Ideal type fit (minimum) 0.80† (0.60)

Prospector deviation ⫺0.21 (0.28) 0.08 (0.42) 0.81 (0.53)
Analyzer deviation ⫺0.02 (0.46) ⫺0.51 (0.68) ⫺0.49 (0.79)
Defender deviation 0.21 (0.28)
Prospector deviation ⫻ ⫺0.80 (0.89)
rate of change
Prospector deviation ⫻ ⫺2.08* (0.95)
uncertainty
Analyzer deviation ⫻ 1.58 (1.65)
rate of change
Analyzer deviation ⫻ 1.05 (1.47)
uncertainty
Environmental rate of 1.19 (0.72) 1.23† (0.73) 1.16 (0.73) 1.23† (0.73) 3.44 (2.61) 1.14 (0.71) ⫺1.64 (2.95) 1.16 (0.72)
change
Environmental uncertainty ⫺0.99 (0.71) ⫺1.01 (0.73) ⫺0.93 (0.72) ⫺1.01 (0.73) ⫺1.00 (0.72) 4.71† (2.64) ⫺0.92 (0.72) ⫺2.81 (2.77)

Constant 2.73* (1.13) 1.96* (0.90) 1.42 (0.90) 0.69 (0.99) 1.16 (1.23) ⫺0.91 (1.49) †
2.28 (1.28) 2.23 (1.47)
Likelihood-ratio ␹2 6.21* 4.91 4.33 4.91 5.75 10.61* 5.27 4.85
Pseudo-R2 0.02 0.02 0.03 0.02 0.02 0.04 0.02 0.02

a
Standard errors are in parentheses; n ⫽ 139; t-tests are one-tailed where predicted, otherwise two-tailed.
†
p ⱕ .10
* p ⱕ .05
420 Academy of Management Journal April

pared with the statistic for a baseline model with only

environmental controls, the pseudo-R2 increases by only
.01 when the ideal type fit measure is included. Accord-
Peer C. Fiss (fiss@marshall.usc.edu) is the McAlister As-
ingly, although the models offer some evidence that fit to
sociate Professor of Business Administration at the Uni-
an ideal type is positive for performance and that fit with
versity of Southern California’s Marshall School of Busi-
the Prospector type is beneficial in uncertain environ-
ness. He received his Ph.D. jointly from the Departments
ments, it would not appear that ideal-type fit is a crucial
of Management & Organization and Sociology at North-
ingredient in attaining high performance.b
western University. His current research interests in-
clude framing and symbolic management, configuration-
al theory and set-theoretic methods, and practice
b diffusion and adaptation.
The results are identical for OLS regression in com-
bination with the uncalibrated performance measure, in-
cluding significance levels.
Copyright of Academy of Management Journal is the property of Academy of Management and its content may
not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written
permission. However, users may print, download, or email articles for individual use.